Skip to content
HypoNum™

Statistical validation

1,656

Analyses listed

1,308

Validated against R

118

Validated against NIST certified values

255

Validated analytically

1

Validated against a published example

0

Validated by an independent reviewer

45

With runnable R code ⟨R⟩

Every analysis window in HypoNum is listed below with the independent references it has been compared against: R, NIST certified datasets, closed-form results, or a published worked example. A tick means the product’s result agreed with the reference under our accuracy policy, which requires agreement to at least four significant digits for probabilities and dispersion statistics and exact equality for counts and decisions. When a comparison disagrees we correct the software; disagreements are not footnoted away. This page is regenerated from the complete comparison suite for each release.

Assessed build: HypoNum 0.1.0.

Explore

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Explain multimodalityOne categorical column separates the two modesdefault settings✓ exact ↗
Explain multimodalityOnly a two-column conjunction identifies the upper modedefault settings✓ exact ↗
Explain multimodalityRetained Poisson fit: exact rate, likelihood, BIC and row allocationsdefault settings✓ Δ 2.2×10-15 (15+ digits) ↗
Explain multimodalityRetained single-Gaussian fit: exact MLE, likelihood, BIC and row allocationsdefault settings✓ Δ 4.9×10-15 (15+ digits) ↗
MultimodalityBimodality coefficient of two separated modesactive methods = BC✓ Δ 3.3×10-16 (15+ digits) ↗
MultimodalityBounded support: dip remains available and automatic KDE signal is refusedmeasurement = continuous; replicates = 999; seed = 17; support = bounded✓ Δ 2.1×10-17 (15+ digits) ↗
MultimodalityCurrent screen: independent dip and bandwidth on two continuous columnsmeasurement = continuous; replicates = 999; seed = 17✓ Δ 3.0×10-17 (15+ digits) ↗
MultimodalityDeclared discrete counts refuse both continuous screen methodsmeasurement = discrete✓ Δ 2.2×10-19 (15+ digits) ↗

Statistics

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Bootstrap estimatesBootstrap estimate 10 percent trimmed meanconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = trimmed_mean; trim = 0.1✓ Δ 1.3×10-15 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate Pearson correlationconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; second = w; seed = 20260928; statistic = pearson✓ Δ 1.1×10-16 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate first quartileconfidence level = 0.95; intervals = percentile; probability = 0.25; resamples = 200; scheme = rows; seed = 20260928; statistic = quantile✓ Δ 1.5×10-15 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate meanconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = mean✓ exact ↗
Bootstrap estimatesBootstrap estimate mean difference B minus Aconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = mean_difference✓ Δ 4.4×10-18 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate mean minus targetconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = mean_minus_target; target = 12✓ exact ↗
Bootstrap estimatesBootstrap estimate mean of paired differencesconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; second = w; seed = 20260928; statistic = paired_mean_difference✓ Δ 8.3×10-18 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate medianconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = median✓ Δ 3.6×10-16 (15+ digits) ↗
Bootstrap estimatesBootstrap estimate standard deviationconfidence level = 0.95; intervals = percentile; resamples = 200; scheme = rows; seed = 20260928; statistic = sd✓ Δ 1.6×10-17 (15+ digits) ↗
Meta-analysisFixeddata = generic; effect = effect; measure = hedges_g; standard error = se; study = study✓ Δ 3.6×10-15 (13 digits) ↗
Meta-analysisRandomconfidence level = 0.99; data = generic; effect = effect; inference = normal; measure = hedges_g; standard error = se; study = study; tau2 = pm✓ Δ 2.5×10-10 (10 digits) ↗
Meta-analysisRandomconfidence level = 0.9; data = continuous; inference = modified_hk; mean1 = mean1; mean2 = mean2; measure = hedges_g; n1 = n1; n2 = n2; sd1 = sd1; sd2 = sd2; study = study; tau2 = dl✓ Δ 6.9×10-15 (14 digits) ↗
Meta-analysisRandomcorrection = 0.5; data = binary; events1 = events1; events2 = events2; inference = hk; measure = log_or; study = study; tau2 = reml; total1 = total1; total2 = total2✓ Δ 1.5×10-14 (13 digits) ↗
Meta-analysisRandomdata = generic; effect = effect; inference = normal; measure = hedges_g; standard error = se; study = study; tau2 = dl✓ Δ 7.1×10-15 (13 digits) ↗

Statistics › Bayesian statistics

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Binomial proportion90% credible interval, event 'fail'confidence level = 0.9; event value = fail; null proportion = 0.5; prior a = 1; prior b = 1✓ Δ 9.7×10-14 (12 digits) ↗
Binomial proportionBeta(1, 1) prior against p = 0.5confidence level = 0.95; event value = pass; null proportion = 0.5; prior a = 1; prior b = 1✓ Δ 2.8×10-13 (12 digits) ↗
Binomial proportionInformative Beta(2, 5) prior against p = 0.3confidence level = 0.95; event value = pass; null proportion = 0.3; prior a = 2; prior b = 5✓ Δ 2.3×10-13 (12 digits) ↗
Binomial proportionThree excluded rowsconfidence level = 0.95; event value = pass; null proportion = 0.5; prior a = 1; prior b = 1✓ Δ 2.4×10-13 (12 digits) ↗
Linear regressionAll models of x1 and x2, Zellner-Siow medium scaleconfidence level = 0.95; hyper a = 3; prior = zellner_siow; prior scale = 0.353553✓ Δ 2.9×10-13 (14 digits) ↗
Linear regressionDeclared models with custom prior weightsconfidence level = 0.95; hyper a = 3; model prior = 0.5, 0.25; stored fits supplied; null model prior = 0.25; prior = zellner_siow; prior scale = 0.353553✓ Δ 2.9×10-13 (14 digits) ↗
Linear regressionHyper-g prior, a = 3confidence level = 0.95; hyper a = 3; prior = hyper_g; prior scale = 0.353553✓ Δ 1.8×10-7 (6 digits) ↗
Linear regressionZellner-Siow unit scale, 90% intervalconfidence level = 0.9; hyper a = 3; prior = zellner_siow; prior scale = 1✓ Δ 6.2×10-13 (13 digits) ↗
Normal mean (known SD)90% credible intervalconfidence level = 0.9; known sd = 1; null value = 0; prior mean = 0; prior sd = 1✓ Δ 2.2×10-13 (12 digits) ↗
Normal mean (known SD)Informative Normal(1, 0.5) prior, known SD 1.2, null 0.5confidence level = 0.95; known sd = 1.2; null value = 0.5; prior mean = 1; prior sd = 0.5✓ Δ 3.4×10-13 (11 digits) ↗
Normal mean (known SD)Normal(0, 1) prior, known SD 1, null 0confidence level = 0.95; known sd = 1; null value = 0; prior mean = 0; prior sd = 1✓ Δ 3.1×10-13 (11 digits) ↗
One-way ANOVAThree-level factor, scale 0.5confidence level = 0.95; prior scale = 0.5✓ Δ 2.2×10-14 (13 digits) ↗
One-way ANOVAThree-level factor, scale 1confidence level = 0.95; prior scale = 1✓ Δ 2.8×10-14 (13 digits) ↗
One-way ANOVATwo-level factor, 90% intervalconfidence level = 0.9; prior scale = 0.5✓ Δ 2.4×10-14 (13 digits) ↗
Pearson correlationStretched Beta(2, 2) prior (width 0.5)confidence level = 0.95; first = x1; prior width = 0.5; second = y✓ Δ 6.8×10-13 (14 digits) ↗
Pearson correlationUniform prior (width 1) on x1 and yconfidence level = 0.95; first = x1; prior width = 1; second = y✓ Δ 1.1×10-12 (14 digits) ↗
Pearson correlationWeak correlation x2 and d, 90% intervalconfidence level = 0.9; first = x2; prior width = 1; second = d✓ Δ 1.4×10-14 (12 digits) ↗
Poisson rate99% credible intervalconfidence level = 0.99; null rate = 2; prior = gamma; prior rate = 1; prior shape = 1✓ Δ 1.4×10-12 (12 digits) ↗
Poisson rateGamma(1, 1) prior, unit exposure, rate 1confidence level = 0.95; null rate = 1; prior = gamma; prior rate = 1; prior shape = 1✓ Δ 2.7×10-13 (13 digits) ↗
Poisson rateGamma(2, 0.5) prior with an exposure columnconfidence level = 0.95; exposure = exposure; null rate = 1.5; prior = gamma; prior rate = 0.5; prior shape = 2✓ Δ 3.3×10-13 (12 digits) ↗
Poisson rateJeffreys prior with an exposure column (posterior only)confidence level = 0.95; exposure = exposure; null rate = 1; prior = jeffreys; prior rate = 1; prior shape = 1✓ Δ 4.0×10-14 (13 digits) ↗
t testsIndependent-samples JZS t, code-point group orderconfidence level = 0.95; design = independent; null value = 0; prior scale = 0.707107✓ Δ 4.2×10-15 (13 digits) ↗
t testsIndependent-samples JZS t, group B firstconfidence level = 0.95; design = independent; null value = 0; prior scale = 0.707107✓ Δ 3.1×10-15 (14 digits) ↗
t testsOne-sample JZS t, default scaleconfidence level = 0.95; design = one_sample; null value = 0; prior scale = 0.707107✓ Δ 3.2×10-14 (13 digits) ↗
t testsOne-sample JZS t, scale 1, null 0.3, 90% intervalconfidence level = 0.9; design = one_sample; null value = 0.3; prior scale = 1✓ Δ 1.7×10-14 (13 digits) ↗
t testsPaired JZS tconfidence level = 0.95; design = paired; null value = 0; prior scale = 0.707107; second = d2✓ Δ 1.8×10-14 (14 digits) ↗

Statistics › Correlation & regression

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Correlation & covarianceCorrelation & covariancedefault settings✓ ↗
Correlation & covarianceCorrelation & covariancemetric = covariance✓ Δ 2.6×10-9 (8 digits) ↗
Correlation & covarianceCorrelation & covariancemetric = pearson✓ Δ 1.1×10-16 (15+ digits) ↗
Correlation & covarianceCorrelation & covariancemetric = spearman✓ Δ 1.1×10-18 (15+ digits) ↗
Correlation & covarianceCorrelation & covarianceunknown option = 1✓ ↗
Correlation & covarianceKendall tau-b exact inferencedeletion = pairwise; inference = exact✓ Δ 2.5×10-17 (9 digits) ↗
Correlation & covarianceKendall tau-b listwise deletiondeletion = listwise; inference = asymptotic✓ Δ 4.9×10-16 (14 digits) ↗
Correlation & covarianceKendall tau-b pairwise deletiondeletion = pairwise; inference = asymptotic✓ Δ 4.9×10-16 (14 digits) ↗
Correlation & covarianceKendall tau-b tie-corrected normaldeletion = pairwise; inference = asymptotic✓ Δ 1.1×10-16 (15+ digits) ↗
Correlation & covarianceMissing and excluded multivariate populationdefault settings✓ Δ 5.1×10-17 (15+ digits) ↗
Correlation & covarianceMultivariatemetric = pearson✓ Δ 1.1×10-16 (15+ digits) ↗
Correlation & covarianceMultivariate optionmetric = covariance✓ Δ 4.5×10-16 (15+ digits) ↗
Correlation & covarianceMultivariate optionmetric = spearman✓ Δ 1.1×10-18 (15+ digits) ↗
Partial correlationPartial correlation one controlcontrols = c1✓ Δ 4.0×10-15 (14 digits) ↗
Partial correlationPartial correlation redundant controlcontrols = c1, c2✓ Δ 7.1×10-15 (13 digits) ↗
Partial correlationPartial correlation two controls with a missing control valuecontrols = c1, c3✓ Δ 6.8×10-15 (14 digits) ↗

Statistics › Correlation & regression › Continuous-response models

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Curve fittingCurve estimation log-scale curves 90 percentconfidence level = 0.9; curves = power, compound, s, growth, exponential✓ Δ 1.5×10-12 (13 digits) ↗
Curve fittingCurve estimation logistic with upper boundconfidence level = 0.95; curves = logistic; upper bound = 50✓ Δ 1.6×10-12 (13 digits) ↗
Curve fittingCurve estimation original-scale curvesconfidence level = 0.95; curves = linear, quadratic, cubic, logarithmic, inverse✓ Δ 7.9×10-11 (12 digits) ↗
Curve fittingCurve estimation sequence with input positionsconfidence level = 0.95; curves = linear, quadratic; sequence = input_position✓ Δ 4.6×10-12 (13 digits) ↗
Curve fittingCurve estimation sequence with retained orderconfidence level = 0.95; curves = linear, quadratic; sequence = retained_order✓ Δ 5.7×10-13 (13 digits) ↗
Curve fittingCurve fittingbad setting = 1; curve = exponential_decay; parameters = a from 4, k from 0.8, c from 1✓ ↗
Curve fittingCurve fittingcurve = asymptotic_growth; parameters = a from 4, k from 0.8, c from 1✓ Δ 5.0×10-10 (9 digits) ↗
Curve fittingCurve fittingcurve = exponential_decay; parameters = a from 4, k from 0.8, c from 1✓ Δ 4.2×10-10 (9 digits) ↗✓ ↗
Curve fittingCurve fittingcurve = exponential_decay; parameters = a from 4, k from 0.8, c from 10001✓ Δ 8.8×10-8 (6 digits) ↗
Curve fittingCurve fittingcurve = four_parameter_logistic; parameters = lo from 1, hi from 5, mid from 4, slope from 2✓ Δ 1.3×10-9 (9 digits) ↗
Curve fittingNIST Bennett5: nonlinear regression Start 1curve = custom; expression = b1 * (b2+x)**(-1/b3); max evaluations = 10000; parameters = b1 from -2000, b2 from 50, b3 from 0.8✓ Δ 6.0×10-7 (10 digits)1 ↗
Curve fittingNIST Bennett5: nonlinear regression Start 2curve = custom; expression = b1 * (b2+x)**(-1/b3); max evaluations = 10000; parameters = b1 from -1500, b2 from 45, b3 from 0.85✓ Δ 1.7×10-6 (10 digits)2 ↗
Curve fittingNIST BoxBOD: nonlinear regression Start 1curve = custom; expression = b1*(1-exp(-b2*x)); max evaluations = 10000; parameters = b1 from 1, b2 from 1✓ Δ 4.4×10-8 (10 digits) ↗
Curve fittingNIST BoxBOD: nonlinear regression Start 2curve = custom; expression = b1*(1-exp(-b2*x)); max evaluations = 10000; parameters = b1 from 100, b2 from 0.75✓ Δ 4.4×10-8 (10 digits)3 ↗
Curve fittingNIST Chwirut1: nonlinear regression Start 1curve = custom; expression = exp(-b1*x)/(b2+b3*x); max evaluations = 10000; parameters = b1 from 0.1, b2 from 0.01, b3 from 0.02✓ Δ 9.4×10-9 (10 digits)4 ↗
Curve fittingNIST Chwirut1: nonlinear regression Start 2curve = custom; expression = exp(-b1*x)/(b2+b3*x); max evaluations = 10000; parameters = b1 from 0.15, b2 from 0.008, b3 from 0.01✓ Δ 9.3×10-9 (10 digits)5 ↗
Curve fittingNIST Chwirut2: nonlinear regression Start 1curve = custom; expression = exp(-b1*x)/(b2+b3*x); max evaluations = 10000; parameters = b1 from 0.1, b2 from 0.01, b3 from 0.02✓ Δ 3.1×10-9 (10 digits)6 ↗
Curve fittingNIST Chwirut2: nonlinear regression Start 2curve = custom; expression = exp(-b1*x)/(b2+b3*x); max evaluations = 10000; parameters = b1 from 0.15, b2 from 0.008, b3 from 0.01✓ Δ 3.1×10-9 (10 digits)7 ↗
Curve fittingNIST DanWood: nonlinear regression Start 1curve = custom; expression = b1*x**b2; max evaluations = 10000; parameters = b1 from 1, b2 from 5✓ Δ 2.4×10-11 (10 digits)8 ↗
Curve fittingNIST DanWood: nonlinear regression Start 2curve = custom; expression = b1*x**b2; max evaluations = 10000; parameters = b1 from 0.7, b2 from 4✓ Δ 2.4×10-11 (10 digits) ↗
Curve fittingNIST ENSO: nonlinear regression Start 1curve = custom; expression = b1 + b2*cos( 2*3.141592653589793*x/12 ) + b3*sin( 2*3.141592653589793*x/12 ) + b5*cos( 2*3.141592653589793*x/b4 ) + b...; max evaluations = 10000; parameters = b1 from 11, b2 from 3, b3 from 0.5, b4 from 40, b5 from -0.7, b6 from -1.3, b7 from 25, b8 from -0.3, b9 from 1.4✓ Δ 2.9×10-9 (10 digits) ↗
Curve fittingNIST ENSO: nonlinear regression Start 2curve = custom; expression = b1 + b2*cos( 2*3.141592653589793*x/12 ) + b3*sin( 2*3.141592653589793*x/12 ) + b5*cos( 2*3.141592653589793*x/b4 ) + b...; max evaluations = 10000; parameters = b1 from 10, b2 from 3, b3 from 0.5, b4 from 44, b5 from -1.5, b6 from 0.5, b7 from 26, b8 from -0.1, b9 from 1.5✓ Δ 2.9×10-9 (10 digits) ↗
Curve fittingNIST Eckerle4: nonlinear regression Start 1curve = custom; expression = (b1/b2) * exp(-0.5*((x-b3)/b2)**2); max evaluations = 10000; parameters = b1 from 1, b2 from 10, b3 from 500✓ Δ 3.9×10-9 (10 digits) ↗
Curve fittingNIST Eckerle4: nonlinear regression Start 2curve = custom; expression = (b1/b2) * exp(-0.5*((x-b3)/b2)**2); max evaluations = 10000; parameters = b1 from 1.5, b2 from 5, b3 from 450✓ Δ 3.9×10-9 (10 digits)9 ↗
Curve fittingNIST Gauss1: nonlinear regression Start 1curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 97, b2 from 0.009, b3 from 100, b4 from 65, b5 from 20, b6 from 70, b7 from 178, b8 from 16.5✓ Δ 3.4×10-9 (10 digits)10 ↗
Curve fittingNIST Gauss1: nonlinear regression Start 2curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 94, b2 from 0.0105, b3 from 99, b4 from 63, b5 from 25, b6 from 71, b7 from 180, b8 from 20✓ Δ 3.4×10-9 (10 digits) ↗
Curve fittingNIST Gauss2: nonlinear regression Start 1curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 96, b2 from 0.009, b3 from 103, b4 from 106, b5 from 18, b6 from 72, b7 from 151, b8 from 18✓ Δ 3.1×10-8 (10 digits) ↗
Curve fittingNIST Gauss2: nonlinear regression Start 2curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 98, b2 from 0.0105, b3 from 103, b4 from 105, b5 from 20, b6 from 73, b7 from 150, b8 from 20✓ Δ 3.1×10-8 (10 digits) ↗
Curve fittingNIST Gauss3: nonlinear regression Start 1curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 94.9, b2 from 0.009, b3 from 90.1, b4 from 113, b5 from 20, b6 from 73.8, b7 from 140, b8 from 20✓ Δ 1.3×10-8 (10 digits) ↗
Curve fittingNIST Gauss3: nonlinear regression Start 2curve = custom; expression = b1*exp( -b2*x ) + b3*exp( -(x-b4)**2 / b5**2 ) + b6*exp( -(x-b7)**2 / b8**2 ); max evaluations = 10000; parameters = b1 from 96, b2 from 0.0096, b3 from 80, b4 from 110, b5 from 25, b6 from 74, b7 from 139, b8 from 25✓ Δ 1.3×10-8 (10 digits)11 ↗
Curve fittingNIST Hahn1: nonlinear regression Start 1curve = custom; expression = (b1+b2*x+b3*x**2+b4*x**3) / (1+b5*x+b6*x**2+b7*x**3); max evaluations = 10000; parameters = b1 from 10, b2 from -1, b3 from 0.05, b4 from -1e-05, b5 from -0.05, b6 from 0.001, b7 from -1e-06✓ Δ 5.9×10-11 (8 digits) ↗
Curve fittingNIST Hahn1: nonlinear regression Start 2curve = custom; expression = (b1+b2*x+b3*x**2+b4*x**3) / (1+b5*x+b6*x**2+b7*x**3); max evaluations = 10000; parameters = b1 from 1, b2 from -0.1, b3 from 0.005, b4 from -1e-06, b5 from -0.005, b6 from 0.0001, b7 from -1e-07✓ Δ 3.9×10-11 (9 digits) ↗
Curve fittingNIST Kirby2: nonlinear regression Start 1curve = custom; expression = (b1 + b2*x + b3*x**2) / (1 + b4*x + b5*x**2); max evaluations = 10000; parameters = b1 from 2, b2 from -0.1, b3 from 0.003, b4 from -0.001, b5 from 1e-05✓ Δ 2.2×10-11 (10 digits)12 ↗
Curve fittingNIST Kirby2: nonlinear regression Start 2curve = custom; expression = (b1 + b2*x + b3*x**2) / (1 + b4*x + b5*x**2); max evaluations = 10000; parameters = b1 from 1.5, b2 from -0.15, b3 from 0.0025, b4 from -0.0015, b5 from 2e-05✓ Δ 2.2×10-11 (10 digits)13 ↗
Curve fittingNIST Lanczos1: nonlinear regression Start 1curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 1.2, b2 from 0.3, b3 from 5.6, b4 from 5.5, b5 from 6.5, b6 from 7.6✓ Δ 3.8×10-11 (6 digits) ↗
Curve fittingNIST Lanczos1: nonlinear regression Start 2curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 0.5, b2 from 0.7, b3 from 3.6, b4 from 4.2, b5 from 4, b6 from 6.3✓ Δ 3.8×10-11 (6 digits)14 ↗
Curve fittingNIST Lanczos2: nonlinear regression Start 1curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 1.2, b2 from 0.3, b3 from 5.6, b4 from 5.5, b5 from 6.5, b6 from 7.6✓ Δ 1.8×10-9 (8 digits) ↗
Curve fittingNIST Lanczos2: nonlinear regression Start 2curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 0.5, b2 from 0.7, b3 from 3.6, b4 from 4.2, b5 from 4, b6 from 6.3✓ Δ 3.7×10-10 (8 digits) ↗
Curve fittingNIST Lanczos3: nonlinear regression Start 1curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 1.2, b2 from 0.3, b3 from 5.6, b4 from 5.5, b5 from 6.5, b6 from 7.6✓ Δ 4.2×10-10 (8 digits)15 ↗
Curve fittingNIST Lanczos3: nonlinear regression Start 2curve = custom; expression = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x); max evaluations = 10000; parameters = b1 from 0.5, b2 from 0.7, b3 from 3.6, b4 from 4.2, b5 from 4, b6 from 6.3✓ Δ 4.9×10-11 (10 digits)16 ↗
Curve fittingNIST MGH09: nonlinear regression Start 1curve = custom; expression = b1*(x**2+x*b2) / (x**2+x*b3+b4); max evaluations = 10000; parameters = b1 from 25, b2 from 39, b3 from 41.5, b4 from 39✓ Δ 4.4×10-12 (10 digits)17 ↗
Curve fittingNIST MGH09: nonlinear regression Start 2curve = custom; expression = b1*(x**2+x*b2) / (x**2+x*b3+b4); max evaluations = 10000; parameters = b1 from 0.25, b2 from 0.39, b3 from 0.415, b4 from 0.39✓ Δ 4.4×10-12 (10 digits)18 ↗
Curve fittingNIST MGH10: nonlinear regression Start 1curve = custom; expression = b1 * exp(b2/(x+b3)); max evaluations = 10000; parameters = b1 from 2, b2 from 400000, b3 from 25000✓ Δ 1.4×10-8 (10 digits) ↗
Curve fittingNIST MGH10: nonlinear regression Start 2curve = custom; expression = b1 * exp(b2/(x+b3)); max evaluations = 10000; parameters = b1 from 0.02, b2 from 4000, b3 from 250✓ Δ 1.6×10-6 (8 digits) ↗
Curve fittingNIST MGH17: nonlinear regression Start 1curve = custom; expression = b1 + b2*exp(-x*b4) + b3*exp(-x*b5); max evaluations = 10000; parameters = b1 from 50, b2 from 150, b3 from -100, b4 from 1, b5 from 2✓ Δ 1.3×10-11 (10 digits) ↗
Curve fittingNIST MGH17: nonlinear regression Start 2curve = custom; expression = b1 + b2*exp(-x*b4) + b3*exp(-x*b5); max evaluations = 10000; parameters = b1 from 0.5, b2 from 1.5, b3 from -1, b4 from 0.01, b5 from 0.02✓ Δ 1.7×10-8 (7 digits)19 ↗
Curve fittingNIST Misra1a: nonlinear regression Start 1curve = custom; expression = b1*(1-exp(-b2*x)); max evaluations = 10000; parameters = b1 from 500, b2 from 0.0001✓ Δ 1.1×10-9 (10 digits)20 ↗
Curve fittingNIST Misra1a: nonlinear regression Start 2curve = custom; expression = b1*(1-exp(-b2*x)); max evaluations = 10000; parameters = b1 from 250, b2 from 0.0005✓ Δ 1.1×10-9 (10 digits)21 ↗
Curve fittingNIST Misra1b: nonlinear regression Start 1curve = custom; expression = b1 * (1-(1+b2*x/2)**(-2)); max evaluations = 10000; parameters = b1 from 500, b2 from 0.0001✓ Δ 3.7×10-9 (10 digits)22 ↗
Curve fittingNIST Misra1b: nonlinear regression Start 2curve = custom; expression = b1 * (1-(1+b2*x/2)**(-2)); max evaluations = 10000; parameters = b1 from 300, b2 from 0.0002✓ Δ 3.7×10-9 (10 digits)23 ↗
Curve fittingNIST Misra1c: nonlinear regression Start 1curve = custom; expression = b1 * (1-(1+2*b2*x)**(-.5)); max evaluations = 10000; parameters = b1 from 500, b2 from 0.0001✓ Δ 3.0×10-9 (10 digits)24 ↗
Curve fittingNIST Misra1c: nonlinear regression Start 2curve = custom; expression = b1 * (1-(1+2*b2*x)**(-.5)); max evaluations = 10000; parameters = b1 from 600, b2 from 0.0002✓ Δ 3.0×10-9 (10 digits)25 ↗
Curve fittingNIST Misra1d: nonlinear regression Start 1curve = custom; expression = b1*b2*x*((1+b2*x)**(-1)); max evaluations = 10000; parameters = b1 from 500, b2 from 0.0001✓ Δ 1.8×10-10 (10 digits)26 ↗
Curve fittingNIST Misra1d: nonlinear regression Start 2curve = custom; expression = b1*b2*x*((1+b2*x)**(-1)); max evaluations = 10000; parameters = b1 from 450, b2 from 0.0003✓ Δ 1.8×10-10 (10 digits)27 ↗
Curve fittingNIST Nelson: nonlinear regression Start 1curve = custom; expression = b1 - b2*x * exp(-b3*x2); max evaluations = 10000; parameters = b1 from 2, b2 from 0.0001, b3 from -0.01✓ Δ 4.5×10-11 (10 digits) ↗
Curve fittingNIST Nelson: nonlinear regression Start 2curve = custom; expression = b1 - b2*x * exp(-b3*x2); max evaluations = 10000; parameters = b1 from 2.5, b2 from 5e-09, b3 from -0.05✓ Δ 4.5×10-11 (10 digits) ↗
Curve fittingNIST NoInt1: fitcurve = custom; expression = b1*x; parameters = b1 from 1✓ Δ 1.5×10-13 (14 digits) ↗
Curve fittingNIST NoInt2: fitcurve = custom; expression = b1*x; parameters = b1 from 1✓ Δ 7.9×10-16 (14 digits) ↗
Curve fittingNIST Rat42: nonlinear regression Start 1curve = custom; expression = b1 / (1+exp(b2-b3*x)); max evaluations = 10000; parameters = b1 from 100, b2 from 1, b3 from 0.1✓ Δ 4.3×10-10 (10 digits)28 ↗
Curve fittingNIST Rat42: nonlinear regression Start 2curve = custom; expression = b1 / (1+exp(b2-b3*x)); max evaluations = 10000; parameters = b1 from 75, b2 from 2.5, b3 from 0.07✓ Δ 4.3×10-10 (10 digits) ↗
Curve fittingNIST Rat43: nonlinear regression Start 1curve = custom; expression = b1 / ((1+exp(b2-b3*x))**(1/b4)); max evaluations = 10000; parameters = b1 from 100, b2 from 10, b3 from 1, b4 from 1✓ Δ 3.7×10-8 (10 digits)29 ↗
Curve fittingNIST Rat43: nonlinear regression Start 2curve = custom; expression = b1 / ((1+exp(b2-b3*x))**(1/b4)); max evaluations = 10000; parameters = b1 from 700, b2 from 5, b3 from 0.75, b4 from 1.3✓ Δ 3.7×10-8 (10 digits)30 ↗
Curve fittingNIST Roszman1: nonlinear regression Start 1curve = custom; expression = b1 - b2*x - arctan(b3/(x-b4))/3.141592653589793; max evaluations = 10000; parameters = b1 from 0.1, b2 from -1e-05, b3 from 1000, b4 from -100✓ Δ 6.7×10-8 (10 digits) ↗
Curve fittingNIST Roszman1: nonlinear regression Start 2curve = custom; expression = b1 - b2*x - arctan(b3/(x-b4))/3.141592653589793; max evaluations = 10000; parameters = b1 from 0.2, b2 from -5e-06, b3 from 1200, b4 from -150✓ Δ 3.2×10-6 (8 digits) ↗
Curve fittingNIST Thurber: nonlinear regression Start 1curve = custom; expression = (b1 + b2*x + b3*x**2 + b4*x**3) / (1 + b5*x + b6*x**2 + b7*x**3); max evaluations = 10000; parameters = b1 from 1000, b2 from 1000, b3 from 400, b4 from 40, b5 from 0.7, b6 from 0.3, b7 from 0.03✓ Δ 2.2×10-4 (6 digits)31 ↗
Curve fittingNIST Thurber: nonlinear regression Start 2curve = custom; expression = (b1 + b2*x + b3*x**2 + b4*x**3) / (1 + b5*x + b6*x**2 + b7*x**3); max evaluations = 10000; parameters = b1 from 1300, b2 from 1500, b3 from 500, b4 from 75, b5 from 1, b6 from 0.4, b7 from 0.05✓ Δ 4.6×10-8 (10 digits)32 ↗
Curve fittingRegression edgecurve = exponential_decay; parameters = a from 9.5, k from 0.8, c from 1✓ Δ 1.5×10-10 (10 digits) ↗
Deming regression - measurement errorDeming regression with lambda 1default settings✓ Δ 2.5×10-13 (13 digits) ↗
Deming regression - measurement errorDeming regression with lambda 2.25comparable units = yes; lambda ratio = 2.25✓ Δ 2.5×10-13 (13 digits) ↗
Deming regression - measurement errorJackknife t intervals on the NCSS ten-pair exampleinterval method = jackknife_t; lambda ratio = 0.25✓ Δ 7.1×10-15 (13 digits) ↗
Deming regression - measurement errorJackknife t intervals with lambda 2.25 on 40 pairsinterval method = jackknife_t; lambda ratio = 2.25✓ Δ 1.3×10-14 (13 digits) ↗
Deming regression - measurement errorMissing and excluded pairs keep their rows without fitted valuesdefault settings✓ Δ 7.2×10-15 (15+ digits) ↗
Deming regression - measurement errorSwapping X and Y with 1/lambda gives the inverse linelambda ratio = 0.444444✓ Δ 8.9×10-15 (13 digits) ↗
Instrumental-variable regression (2SLS)STAT REG TWO STAGE 001confidence level = 0.95; covariance = classical; endogenous = x; exogenous = w; instruments = z1, z2✓ Δ 2.3×10-13 (13 digits) ↗
Instrumental-variable regression (2SLS)STAT REG TWO STAGE 002confidence level = 0.95; covariance = hc1; endogenous = x; exogenous = w; instruments = z1, z2✓ Δ 1.2×10-13 (13 digits) ↗
Instrumental-variable regression (2SLS)STAT REG TWO STAGE 003confidence level = 0.9; covariance = classical; endogenous = x; instruments = z1✓ Δ 1.7×10-13 (13 digits) ↗
Instrumental-variable regression (2SLS)STAT REG TWO STAGE 004confidence level = 0.99; covariance = hc1; endogenous = x; exogenous = w; instruments = z1, z2✓ Δ 8.5×10-14 (13 digits) ↗
Linear regressionBackwardblocks = x1, x2, x3, x4, x5; casewise = none; casewise threshold = 3; confidence level = 0.95; criterion = bic✓ Δ 1.1×10-13 (13 digits) ↗
Linear regressionBest subsetscriterion = aic✓ exact ↗
Linear regressionBest subsetscriterion = bic✓ exact ↗
Linear regressionEnterblocks = x1, x2, x3, x4; casewise = all; casewise threshold = 3; confidence level = 0.95; criterion = aic✓ Δ 1.1×10-13 (12 digits) ↗
Linear regressionExtreme scaledefault settings✓ Δ 9.8×10-6 (15+ digits) ↗
Linear regressionExtreme scaledegree = 2; interactions = no; missing = drop; model type = linear; split percent = 100✓ Δ 4.8×10-17 (15+ digits) ↗
Linear regressionFit axisalgorithm = lasso; cv folds = 3✓ exact ↗
Linear regressionFit axisalgorithm = ridge; cv folds = 3✓ exact ↗
Linear regressionFit axisalpha level = 0.1✓ Δ 2.1×10-16 (15+ digits) ↗
Linear regressionFit axiscovariance = cluster✓ Δ 1.2×10-17 (14 digits) ↗
Linear regressionFit axiscovariance = hc3✓ Δ 1.7×10-19 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = drop; model type = exponential; split percent = 100✓ Δ 4.8×10-17 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = drop; model type = logarithmic; split percent = 100✓ Δ 2.2×10-16 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = drop; model type = polynomial; split percent = 100✓ Δ 2.0×10-13 (14 digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = drop; model type = power; split percent = 100✓ Δ 2.2×10-16 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = knn; model type = linear; split percent = 100✓ Δ 2.7×10-16 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = no; missing = neighbor_average; model type = linear; split percent = 100✓ Δ 2.7×10-16 (15+ digits) ↗
Linear regressionFit axisdegree = 2; interactions = yes; missing = drop; model type = linear; split percent = 100✓ Δ 2.7×10-16 (15+ digits) ↗
Linear regressionFit axisdegree = 3; model type = polynomial✓ Δ 5.1×10-16 (13 digits) ↗
Linear regressionFit axisinteractions = yes✓ Δ 1.0×10-16 (12 digits) ↗
Linear regressionFit axismissing = knn✓ Δ 1.8×10-16 (15+ digits) ↗
Linear regressionFit axismissing = neighbor_average✓ Δ 1.8×10-16 (15+ digits) ↗
Linear regressionFit axismodel type = exponential✓ Δ 2.5×10-17 (15+ digits) ↗
Linear regressionFit axismodel type = logarithmic✓ Δ 3.6×10-17 (15+ digits) ↗
Linear regressionFit axismodel type = polynomial✓ Δ 2.3×10-16 (13 digits) ↗
Linear regressionFit axismodel type = power✓ Δ 5.6×10-17 (15+ digits) ↗
Linear regressionFit axissplit percent = 80✓ exact ↗
Linear regressionForwardblocks = x1, x2, x3, x4, x5; casewise = none; casewise threshold = 3; confidence level = 0.95; criterion = aic✓ Δ 1.1×10-13 (13 digits) ↗
Linear regressionLinear regressionalgorithm = lasso; cv folds = 3; imputation = drop; seed = 7; test fraction = 0.25✓ exact ↗
Linear regressionLinear regressionalgorithm = ols; bad setting = 1; imputation = drop; seed = 0; test fraction = 0.25✓ ↗
Linear regressionLinear regressionalgorithm = ols; cv folds = 3; imputation = drop✓ exact ↗
Linear regressionLinear regressionalgorithm = ols; cv folds = 3; imputation = knn; seed = 2✓ exact ↗
Linear regressionLinear regressionalgorithm = ols; cv folds = 3; imputation = mean; seed = 2✓ exact ↗
Linear regressionLinear regressionalgorithm = ols; imputation = drop; seed = 0; test fraction = 0.2✓ exact ↗
Linear regressionLinear regressionalgorithm = ols; imputation = drop; seed = 0; test fraction = 0.25✓ exact ↗✓ ↗
Linear regressionLinear regressionalgorithm = ridge; cv folds = 3; imputation = drop; seed = 7; test fraction = 0.25✓ exact ↗
Linear regressionLinear regressionalpha level = 2✓ ↗
Linear regressionLinear regressionbad setting = 1✓ ↗
Linear regressionLinear regressiondefault settings✓ Δ 3.2×10-12 (11 digits) ↗✓ ↗
Linear regressionLinear regressiondegree = 2; interactions = no; missing = bad; model type = linear; split percent = 100✓ ↗
Linear regressionLinear regressiondegree = 2; interactions = no; missing = drop; model type = linear; split percent = 100✓ Δ 9.2×10-13 (13 digits) ↗✓ ↗
Linear regressionLinear regressionmodel type = logarithmic✓ ↗
Linear regressionMissing and excludeddefault settings✓ Δ 7.1×10-16 (15+ digits) ↗
Linear regressionMissing and excludeddegree = 2; interactions = no; missing = drop; model type = linear; split percent = 100✓ Δ 2.3×10-16 (15+ digits) ↗
Linear regressionNIST AtmWtAg: fitalgorithm = ols; model type = linear✓ Δ 2.5×10-14 (14 digits) ↗
Linear regressionNIST Filip: fitalgorithm = ols; degree = 10; model type = polynomial✓ Δ 3.2×10-11 (13 digits) ↗
Linear regressionNIST Longley: fitalgorithm = ols; model type = linear✓ Δ 1.1×10-9 (13 digits)33 ↗
Linear regressionNIST Norris: fitalgorithm = ols; model type = linear✓ Δ 1.1×10-7 (13 digits) ↗
Linear regressionNIST Pontius: fitalgorithm = ols; degree = 2; model type = polynomial✓ Δ 1.6×10-6 (13 digits)34 ↗
Linear regressionNIST SiRstv: fitalgorithm = ols; model type = linear✓ Δ 1.0×10-13 (13 digits) ↗
Linear regressionNIST SmLs01: fitalgorithm = ols; model type = linear✓ Δ 3.6×10-15 (15+ digits) ↗
Linear regressionNIST SmLs02: fitalgorithm = ols; model type = linear✓ Δ 5.7×10-14 (15+ digits) ↗
Linear regressionNIST SmLs03: fitalgorithm = ols; model type = linear✓ Δ 9.1×10-13 (15+ digits) ↗
Linear regressionNIST SmLs04: fitalgorithm = ols; model type = linear✓ Δ 3.6×10-15 (15+ digits) ↗
Linear regressionNIST SmLs05: fitalgorithm = ols; model type = linear✓ Δ 5.7×10-14 (15+ digits) ↗
Linear regressionNIST SmLs06: fitalgorithm = ols; model type = linear✓ Δ 4.5×10-13 (15+ digits) ↗
Linear regressionNIST SmLs07: fitalgorithm = ols; model type = linear✓ Δ 3.6×10-15 (15+ digits) ↗
Linear regressionNIST SmLs08: fitalgorithm = ols; model type = linear✓ Δ 5.7×10-14 (15+ digits) ↗
Linear regressionNIST SmLs09: fitalgorithm = ols; model type = linear✓ Δ 4.5×10-13 (15+ digits) ↗
Linear regressionNIST Wampler1: fitalgorithm = ols; degree = 5; model type = polynomial✓ Δ 1.5×10-11 (10 digits) ↗
Linear regressionNIST Wampler2: fitalgorithm = ols; degree = 5; model type = polynomial✓ Δ 6.4×10-16 (13 digits) ↗
Linear regressionNIST Wampler3: fitalgorithm = ols; degree = 5; model type = polynomial✓ Δ 2.8×10-7 (10 digits) ↗
Linear regressionNIST Wampler4: fitalgorithm = ols; degree = 5; model type = polynomial✓ Δ 1.2×10-4 (9 digits) ↗
Linear regressionNIST Wampler5: fitalgorithm = ols; degree = 5; model type = polynomial✓ Δ 5.9×10-6 (7 digits) ↗
Linear regressionRegression edgealgorithm = ols; imputation = drop; seed = 0; test fraction = 0.5✓ ↗
Linear regressionRegression edgedirection = forward✓ exact ↗
Linear regressionRegression settingdefault settings✓ exact ↗
Linear regressionRegression settingdegree = 2; interactions = no; missing = drop; model type = linear; split percent = 80✓ exact ↗
Linear regressionRegression settingdegree = 3; interactions = no; missing = drop; model type = polynomial; split percent = 100✓ Δ 1.8×10-11 (13 digits) ↗
Linear regressionSTAT REG OLS 001default settings✓ Δ 4.7×10-16 (15+ digits) ↗
Linear regressionSTAT REG VIF 001degree = 2; interactions = no; missing = drop; model type = linear; split percent = 100✓ Δ 2.7×10-16 (15+ digits) ↗
Linear regressionStepwiseblocks = x1, x2, x3, x4, x5; casewise = outliers; casewise threshold = 1.5; confidence level = 0.9; criterion = aic✓ Δ 1.1×10-13 (13 digits) ↗
Linear regressionStepwisecriterion = aic; direction = backward✓ exact ↗
Linear regressionStepwisecriterion = aic; direction = both✓ exact ↗
Linear regressionStepwisecriterion = aic; direction = forward✓ exact ↗
Linear regressionStepwisecriterion = bic; direction = backward✓ exact ↗
Linear regressionStepwisecriterion = bic; direction = both✓ exact ↗
Linear regressionStepwisecriterion = bic; direction = forward✓ exact ↗
Linear regressionTrue minimum sampledefault settings✓ Δ 8.8×10-16 (15+ digits) ↗
Partial least squares regressionPartial least squares regressionbad setting = 1; cv folds = 3; scale = no✓ ↗
Partial least squares regressionPartial least squares regressioncv folds = 2; scale = no; seed = 0; test fraction = 0.25✓ exact ↗
Partial least squares regressionPartial least squares regressioncv folds = 3; imputation = drop; scale = no✓ exact ↗
Partial least squares regressionPartial least squares regressioncv folds = 3; imputation = mean; scale = no✓ exact ↗
Partial least squares regressionPartial least squares regressioncv folds = 3; scale = no✓ exact ↗✓ ↗
Partial least squares regressionPartial least squares regressioncv folds = 3; scale = yes; seed = 3✓ exact ↗
Partial least squares regressionPartial least squares regressionstored fit supplied✓ Δ 6.3×10-10 (10 digits) ↗✓ ↗
Partial least squares regressionPartial least squares regressionstored fit supplied; bad setting = 1✓ ↗
Partial least squares regressionRegression edgecv folds = 2; scale = no; seed = 0; test fraction = 0.25✓ ↗
Partial least squares regressionTrue minimum samplecv folds = 2; scale = no; seed = 0; test fraction = 0.2✓ exact ↗
Quantile regressionSTAT REG QUANTILE REGRESSION 001confidence level = 0.95; dependence = independent; held out evaluation = no; quantiles = 0.5; seed = 0✓ Δ 4.0×10-15 (14 digits) ↗
Quantile regressionSTAT REG QUANTILE REGRESSION 002confidence level = 0.9; dependence = independent; held out evaluation = no; quantiles = 0.75, 0.25; seed = 3✓ Δ 9.1×10-15 (14 digits) ↗
Quantile regressionSTAT REG QUANTILE REGRESSION 003confidence level = 0.95; dependence = independent; held out evaluation = no; quantiles = 0.5; seed = 0✓ Δ 7.4×10-15 (15+ digits) ↗
Quantile regressionSTAT REG QUANTILE REGRESSION 004confidence level = 0.95; dependence = independent; held out evaluation = no; quantiles = 0.5; seed = 0✓ Δ 4.9×10-16 (15+ digits) ↗
Regression with estimated variance weightsSTAT REG WEIGHT ESTIMATION 001confidence level = 0.95; power from = -2; power step = 0.5; power to = 2✓ Δ 2.9×10-14 (14 digits) ↗
Regression with estimated variance weightsSTAT REG WEIGHT ESTIMATION 002confidence level = 0.9; power from = -1; power step = 0.25; power to = 2✓ Δ 8.3×10-14 (14 digits) ↗
Regression with estimated variance weightsSTAT REG WEIGHT ESTIMATION 003confidence level = 0.95; power from = -2; power step = 1; power to = 0✓ Δ 3.6×10-14 (13 digits) ↗
Regression with estimated variance weightsSTAT REG WEIGHT ESTIMATION 004confidence level = 0.99; power from = -1; power step = 0.5; power to = 2✓ Δ 1.6×10-13 (13 digits) ↗

Statistics › Correlation & regression › Generalized & categorical-response models

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Binary regressionBinary regressionevent value = event✓ ↗
Binary regressionBinary regressionevent value = event; threshold = 1✓ ↗
Binary regressionExtreme scaleevent value = event✓ Δ 1.1×10-13 (15+ digits) ↗
Binary regressionMissing and excludedevent value = event✓ Δ 8.7×10-17 (14 digits) ↗
Binary regressionRegression completionevent value = event✓ Δ 3.3×10-16 (15+ digits) ↗
Binary regressionRegression completionevent value = event; threshold = 0.0835259✓ Δ 3.4×10-19 (15+ digits) ↗
Binary regressionRegression completionevent value = event; threshold = 0.0840219✓ Δ 2.3×10-19 (15+ digits) ↗
Binary regressionRegression settingconfidence level = 0.9; event value = event; threshold = 0.3✓ Δ 2.2×10-16 (15+ digits) ↗
Binary regressionRegression settingevent value = event✓ Δ 8.7×10-16 ↗
Binary regressionSTAT REG BINARY 001event value = event✓ Δ 3.0×10-16 (15+ digits) ↗
Multinomial logistic regressionExtreme scalereference value = A✓ Δ 5.6×10-7 (8 digits) ↗
Multinomial logistic regressionMissing and excludedreference value = A✓ Δ 1.0×10-10 (8 digits) ↗
Multinomial logistic regressionMultinomial logistic regressionconfidence level = 1✓ ↗
Multinomial logistic regressionMultinomial logistic regressiondefault settings✓ ↗
Multinomial logistic regressionMultinomial logistic regressionreference value = A✓ ↗
Multinomial logistic regressionRegression settingconfidence level = 0.9; reference value = C✓ Δ 1.7×10-9 (8 digits) ↗
Multinomial logistic regressionSTAT REG NOMINAL 001reference value = A✓ Δ 7.9×10-11 (6 digits) ↗
Negative binomial regressionExtreme scalecount = nb_count; exposure = exposure✓ Δ 1.1×10-10 (13 digits) ↗
Negative binomial regressionMissing and excludedcount = nb_count; exposure = exposure✓ Δ 2.6×10-14 (12 digits) ↗
Negative binomial regressionNegative binomial regressionconfidence level = 1; count = nb_count✓ ↗
Negative binomial regressionNegative binomial regressioncount = nb_count✓ Δ 4.0×10-13 (12 digits) ↗✓ ↗
Negative binomial regressionNegative binomial regressioncount = nb_count; exposure = exposure✓ ↗
Negative binomial regressionRegression settingconfidence level = 0.9; count = nb_count; exposure = exposure; max rows = 100✓ Δ 2.1×10-13 (12 digits) ↗
Negative binomial regressionRegression settingcount = nb_count; exposure = exposure✓ Δ 2.6×10-14 (12 digits) ↗
Negative binomial regressionSTAT REG NB 001count = nb_count; exposure = exposure✓ Δ 2.3×10-12 (10 digits) ↗
Ordinal logistic regressionExtreme scalelevel order = low, middle, high✓ Δ 1.3×10-4 (6 digits) ↗
Ordinal logistic regressionMissing and excludedlevel order = low, middle, high✓ Δ 1.6×10-8 (6 digits) ↗
Ordinal logistic regressionOrdinal logistic regressionconfidence level = 1✓ ↗
Ordinal logistic regressionOrdinal logistic regressiondefault settings✓ ↗
Ordinal logistic regressionOrdinal logistic regressionlevel order = low, middle, high✓ Δ 2.5×10-7 (7 digits) ↗
Ordinal logistic regressionRegression settingconfidence level = 0.9; level order = low, middle, high✓ Δ 1.1×10-7 (6 digits) ↗
Ordinal logistic regressionSTAT REG ORDINAL 001level order = low, middle, high✓ Δ 1.3×10-8 (6 digits) ↗
Other GLM familiesBinomialcomparisons = holm; confidence level = 0.95; covariance = hc0; covariates = x; family = binomial; link = probit; lr tests = no; scale = default; trials = trials✓ Δ 4.1×10-8 (7 digits) ↗
Other GLM familiesGammacomparisons = holm; confidence level = 0.95; covariance = model; covariates = x; family = gamma; link = canonical; lr tests = no; scale = default✓ Δ 3.8×10-12 (11 digits) ↗
Other GLM familiesGaussiancomparisons = holm; confidence level = 0.95; covariance = model; covariates = x; family = gaussian; link = canonical; lr tests = no; scale = default✓ Δ 9.2×10-15 (14 digits) ↗
Other GLM familiesNegative binomialcomparisons = holm; confidence level = 0.95; covariance = model; covariates = x; family = negative_binomial; link = canonical; lr tests = no; scale = default; theta = estimate✓ Δ 2.0×10-8 (7 digits) ↗
Other GLM familiesPoissoncomparisons = holm; confidence level = 0.9; covariance = model; covariates = x; family = poisson; link = canonical; lr tests = yes; offset = log_exposure; scale = default✓ Δ 3.0×10-13 (12 digits) ↗
Poisson regressionExtreme scaledefault settings✓ Δ 1.4×10-7 (10 digits) ↗
Poisson regressionMissing and excludeddefault settings✓ Δ 2.1×10-11 (9 digits) ↗
Poisson regressionPoisson regressionconfidence level = 1✓ ↗
Poisson regressionPoisson regressiondefault settings✓ Δ 8.2×10-9 (9 digits) ↗✓ ↗
Poisson regressionRegression settingconfidence level = 0.9✓ Δ 2.3×10-7 (6 digits) ↗
Poisson regressionRegression settingdefault settings✓ Δ 2.5×10-12 (10 digits) ↗
Poisson regressionSTAT REG POISSON 001default settings✓ Δ 1.4×10-11 (10 digits) ↗

Statistics › Correlation & regression › Predictive methods

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Decision treesCHAID root split on a binary nominal predictormax depth = 3; min child = 1; min parent = 2; target = late; target kind = categorical✓ Δ 1.3×10-23 (15+ digits) ↗
Decision treesCHAID root split on a three-category nominal predictormax depth = 3; min child = 1; min parent = 2; target = grade; target kind = categorical✓ Δ 2.4×10-27 (15+ digits) ↗
Decision treesCHAID root split on an ordered three-category predictormax depth = 3; min child = 1; min parent = 2; ordinal = machine; target = grade; target kind = categorical✓ Δ 2.4×10-27 (15+ digits) ↗
Decision treesCHAID root split with Kass merge/re-splitalgorithm = kass_resplit; max depth = 3; min child = 1; min parent = 2; resplit alpha = 0.01; target = grade; target kind = categorical✓ Δ 2.4×10-27 (15+ digits) ↗
Decision treesDecision treesbad setting = 1; cv folds = 2; depth grid = 1; max depth = 1; min leaf grid = 1✓ ↗
Decision treesDecision treesbad setting = 1; cv folds = 2; max depth = 1; min samples leaf = 1; positive class = B; fixed train/test split supplied✓ ↗
Decision treesDecision treescv folds = 2; depth grid = 1; max depth = 1; min leaf grid = 1✓ exact ↗✓ ↗
Decision treesDecision treescv folds = 2; depth grid = 1; max depth = 1; min leaf grid = 1; seed = 7✓ exact ↗
Decision treesDecision treescv folds = 2; max depth = 1; min samples leaf = 1; positive class = B✓ exact ↗✓ ↗
Decision treesDecision treescv folds = 2; max depth = 1; min samples leaf = 1; positive class = B; fixed train/test split supplied✓ exact ↗
Decision treesDecision treesstored fit supplied✓ exact ↗✓ ↗
Decision treesDecision treesstored fit supplied; bad setting = 1✓ ↗
Decision treesLarge-offset retained tree scoringstored fit supplied✓ exact ↗
Decision treesQUEST root split between two point-mass classes, no pruningmax depth = 5; min child = 1; min parent = 2; prune = no; target = class✓ exact ↗
Decision treesQUEST root split kept by cost-complexity pruningcv folds = 5; max depth = 5; min child = 1; min parent = 2; prune = yes; target = class✓ exact ↗
Decision treesQUEST root split with equal priorsmax depth = 5; min child = 1; min parent = 2; priors = equal; prune = no; target = class✓ exact ↗
Decision treesRegression edgecv folds = 2; depth grid = 1; max depth = 1; min leaf grid = 1✓ exact ↗
Decision treesRegression edgecv folds = 2; max depth = 1; min samples leaf = 1; positive class = B✓ ↗
Decision treesTrue minimum samplecv folds = 2; depth grid = 1; max depth = 1; min leaf grid = 1✓ exact ↗
Decision treesTrue minimum samplecv folds = 2; max depth = 1; min samples leaf = 1; positive class = B; fixed train/test split supplied✓ exact ↗
Decision treesTrue minimum samplestored fit supplied✓ exact ↗
Discriminant analysisDiscriminant analysisstored fit supplied✓ Δ 1.0×10-17 (14 digits) ↗✓ ↗
Discriminant analysisDiscriminant analysisstored fit supplied; bad setting = 1✓ ↗
Discriminant analysisLDAbad setting = 1; prior mode = equal; standardize = no✓ ↗
Discriminant analysisLDAcustom priors = 0.7, 0.3; prior mode = custom; standardize = yes✓ Δ 4.4×10-18 (15+ digits) ↗
Discriminant analysisLDAimputation = drop; prior mode = equal; standardize = no✓ exact ↗
Discriminant analysisLDAimputation = mean; prior mode = equal; standardize = no✓ exact ↗
Discriminant analysisLDAprior mode = equal; seed = 0; standardize = no; test fraction = 0.25✓ exact ↗✓ ↗
Discriminant analysisLDAprior mode = equal; standardize = no✓ exact ↗✓ ↗
Discriminant analysisLDAprior mode = observed; standardize = yes✓ exact ↗
Discriminant analysisQDAimputation = mean; prior mode = equal; standardize = no✓ exact ↗
Discriminant analysisQDAprior mode = equal; standardize = no✓ exact ↗
Elastic-net logistic regressionSTAT REG PENALIZED LOGISTIC 001event = fail; independence declared = yes; seed = 20260923; threshold = 0.5✓ Δ 3.5×10-8 (8 digits) ↗
Nearest-neighbour classification & regressionNearest-neighbour classification of separable clustersdistance = euclidean; features = f1, f2; k values = 1, 3; target = outcome; task = classification✓ exact ↗
Nearest-neighbour classification & regressionNearest-neighbour classification with city-block distancedistance = cityblock; features = f1, f2; k values = 1, 3; target = outcome; task = classification✓ exact ↗
Nearest-neighbour classification & regressionNearest-neighbour regression with constant within-cluster targetsdistance = euclidean; features = f1, f2; k values = 1, 3; target = hours; task = regression✓ exact ↗
Random forest regressionRegression forest refuses dependent rowsdependence = clustered; seed = 7✓ ↗
Random forest regressionRegression forest study design on independent rowsdependence = independent; seed = 7✓ Δ 1.1×10-18 (15+ digits) ↗

Statistics › Distributions & randomness

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Fit & assess distributionsA constant column has no distributioncandidates = normal; origin = 0✓ ↗
Fit & assess distributionsAll eight candidates on positive unimodal datacandidates = normal, student_t, laplace, logistic, lognormal, gamma, weibull, exponential; origin = 0✓ Δ 1.8×10-10 (10 digits) ↗
Fit & assess distributionsAn unknown candidate is refusedcandidates = normal, cauchy; origin = 0✓ ↗
Fit & assess distributionsAnderson darlingdefault settings✓ Δ 1.8×10-15 (13 digits) ↗
Fit & assess distributionsAnderson-Darling statistic of the normal fitcalibration = fixed; candidates = normal; mc samples = 99; origin = 0; seed = 17✓ Δ 2.8×10-16 (15+ digits) ↗
Fit & assess distributionsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Fit & assess distributionsCandidate ranking by BIC then AICcandidates = normal, laplace, logistic, lognormal, gamma, weibull, exponential; origin = 0✓ Δ 3.6×10-16 (15+ digits) ↗
Fit & assess distributionsChi square associationdefault settings✓ exact ↗
Fit & assess distributionsChi square gofdefault settings✓ exact ↗
Fit & assess distributionsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Fit & assess distributionsExponential with a value at its origincandidates = exponential; origin = 0✓ Δ 7.6×10-15 (15+ digits) ↗
Fit & assess distributionsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Fit & assess distributionsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Fit & assess distributionsFixed-origin positive families on 1 to 8candidates = lognormal, gamma, weibull, exponential; origin = 0✓ Δ 7.5×10-15 (15+ digits) ↗
Fit & assess distributionsGrubbsdefault settings✓ Δ 1.4×10-15 (12 digits) ↗
Fit & assess distributionsHybrid fixed-origin lognormal normality testcalibration = hybrid; candidates = lognormal; origin = 2.5✓ Δ 4.1×10-16 (15+ digits) ↗
Fit & assess distributionsHybrid normality test with separate sample-SD estimatorscalibration = hybrid; candidates = normal; origin = 0✓ Δ 1.2×10-14 (13 digits) ↗
Fit & assess distributionsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Fit & assess distributionsLarge offset with a small spreadcandidates = normal, laplace; origin = 0✓ Δ 3.0×10-7 (8 digits) ↗
Fit & assess distributionsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Fit & assess distributionsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Fit & assess distributionsMissing values and excluded rowscandidates = normal, laplace; origin = 0✓ Δ 1.4×10-14 (15+ digits) ↗
Fit & assess distributionsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Fit & assess distributionsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Fit & assess distributionsOne sample talpha = 0✓ ↗
Fit & assess distributionsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Fit & assess distributionsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Fit & assess distributionsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Fit & assess distributionsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Fit & assess distributionsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Fit & assess distributionsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Fit & assess distributionsOne sample tdefault settings✓ ↗
Fit & assess distributionsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Fit & assess distributionsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Fit & assess distributionsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Fit & assess distributionsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Fit & assess distributionsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Fit & assess distributionsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Fit & assess distributionsPositive families at a declared origin of 2.5candidates = lognormal, gamma, weibull, exponential; origin = 2.5✓ Δ 1.4×10-14 (15+ digits) ↗
Fit & assess distributionsPositive families skipped for a value at the origincandidates = normal, lognormal, gamma; origin = 0✓ Δ 3.9×10-15 (15+ digits) ↗
Fit & assess distributionsSeven values are below the minimumcandidates = normal; origin = 0✓ ↗
Fit & assess distributionsShapiro wilkdefault settings✓ Δ 2.2×10-16 (15+ digits) ↗
Fit & assess distributionsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Fit & assess distributionsSmallest accepted sample of eightcandidates = normal, exponential; origin = 0✓ Δ 3.7×10-15 (15+ digits) ↗
Fit & assess distributionsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Fit & assess distributionsUnrestricted Student t is refusedcandidates = student_t; origin = 0✓ ↗
Fit & assess distributionsUnsupporteddefault settings✓ ↗
Fit & assess distributionsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Fit & assess distributionsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Fit & assess distributionsWilcoxondefault settings✓ exact ↗
Fit & assess distributionsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Outlier test - GrubbsAnderson darlingdefault settings✓ Δ 1.8×10-15 (13 digits) ↗
Outlier test - GrubbsGrubbsdefault settings✓ Δ 1.4×10-15 (12 digits) ↗
Outlier test - GrubbsShapiro wilkdefault settings✓ Δ 2.2×10-16 (15+ digits) ↗
Runs test for sequence randomnessFriedmanalternative = less✓ ↗
Runs test for sequence randomnessFriedmandefault settings✓ Δ 3.1×10-18 (15+ digits) ↗✓ ↗
Runs test for sequence randomnessFriedmanmax blocks = 10; max rows = 100✓ Δ 2.3×10-18 (15+ digits) ↗
Runs test for sequence randomnessRunsalternative = greater; inference = exact; reference mode = fixed; reference value = 6✓ Δ 1.4×10-18 (15+ digits) ↗
Runs test for sequence randomnessRunsalternative = less; inference = exact; reference mode = fixed; reference value = 6✓ Δ 1.0×10-19 (15+ digits) ↗
Runs test for sequence randomnessRunsalternative = two-sided; inference = asymptotic; max exact n = 20; reference mode = median✓ Δ 1.7×10-18 (15+ digits) ↗
Runs test for sequence randomnessRunsdefault settings✓ ↗
Runs test for sequence randomnessRunsinference = asymptotic✓ Δ 1.7×10-18 (15+ digits) ↗
Runs test for sequence randomnessRunsinference = asymptotic; reference mode = median✓ ↗

Statistics › Group comparisons & ANOVA

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Analysis of means (ANOM)Binomialalpha = 0.05; analysis = binomial; events = events; trials = trials✓ Δ 8.0×10-14 (13 digits) ↗
Analysis of means (ANOM)Normal one wayalpha = 0.01; analysis = normal_one_way✓ Δ 5.3×10-14 (13 digits) ↗
Analysis of means (ANOM)Normal one wayalpha = 0.05; analysis = normal_one_way✓ Δ 3.9×10-14 (13 digits) ↗
Analysis of means (ANOM)Normal two wayalpha = 0.05; analysis = normal_two_way✓ Δ 1.1×10-14 (13 digits) ↗
Analysis of means (ANOM)Poissonalpha = 0.05; analysis = poisson; events = counts; exposure = exposure; exposure unit = hour✓ Δ 9.3×10-16 (15+ digits) ↗
Compare group distributionsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Compare group distributionsChi square associationdefault settings✓ exact ↗
Compare group distributionsChi square gofdefault settings✓ exact ↗
Compare group distributionsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Compare group distributionsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Compare group distributionsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Compare group distributionsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Compare group distributionsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Compare group distributionsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Compare group distributionsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Compare group distributionsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Compare group distributionsOne sample talpha = 0✓ ↗
Compare group distributionsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Compare group distributionsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Compare group distributionsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Compare group distributionsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Compare group distributionsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Compare group distributionsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Compare group distributionsOne sample tdefault settings✓ ↗
Compare group distributionsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Compare group distributionsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Compare group distributionsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Compare group distributionsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Compare group distributionsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Compare group distributionsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Compare group distributionsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Compare group distributionsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Compare group distributionsUnsupporteddefault settings✓ ↗
Compare group distributionsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Compare group distributionsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Compare group distributionsWilcoxondefault settings✓ exact ↗
Compare group distributionsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Equivalence testsAdditional configuration coveragealpha = 0.1; lower bound = -2; max cells = 1000; max rows = 100; mode = one_sample_difference; target = 6; upper bound = 2✓ Δ 1.1×10-16 (14 digits) ↗
Equivalence testsAdditional configuration coveragealpha = 0.1; lower bound = -3; max cells = 1000; max rows = 100; upper bound = 3✓ Δ 8.8×10-16 (14 digits) ↗
Equivalence testsBelow minimum: One response in each group is below the Welch TOST minimum.lower bound = -2; upper bound = 2✓ ↗
Equivalence testsBelow minimum: One retained response is below the one-sample TOST minimum.lower bound = -2; mode = one_sample_difference; target = 0; upper bound = 2✓ ↗
Equivalence testsDecision thresholdalpha = 0.1; lower bound = -2.7781; upper bound = -0.371895✓ Δ 5.0×10-16 (14 digits) ↗
Equivalence testsDecision thresholdalpha = 0.1; lower bound = -2.80241; upper bound = -0.34759✓ Δ 5.8×10-16 (14 digits) ↗
Equivalence testsDecision thresholdalpha = 0.1; lower bound = 0.160571; mode = one_sample_difference; target = 6; upper bound = 1.46443✓ Δ 8.3×10-16 (14 digits) ↗
Equivalence testsDecision thresholdalpha = 0.1; lower bound = 0.17348; mode = one_sample_difference; target = 6; upper bound = 1.45152✓ Δ 2.4×10-16 (14 digits) ↗
Equivalence testsEquivalence testsalpha = 0.1; lower bound = -3; max rows = 100; mode = one_sample_difference; target = 1.5; upper bound = 3✓ Δ 1.4×10-17 (15+ digits) ↗
Equivalence testsEquivalence testsalpha = 0.1; lower bound = -3; max rows = 100; upper bound = 3✓ Δ 7.3×10-18 (15+ digits) ↗
Equivalence testsEquivalence testslower bound = -1; mode = one_sample_difference; target = 1; upper bound = 1✓ ↗
Equivalence testsEquivalence testslower bound = -1; upper bound = 1✓ ↗
Equivalence testsEquivalence testslower bound = -2; mode = paired_difference; upper bound = 2✓ Δ 2.0×10-18 (13 digits) ↗✓ ↗
Equivalence testsEquivalence testslower bound = -2; upper bound = 2✓ ↗
Equivalence testsEquivalence testslower bound = 0.8; mode = paired_log_ratio; upper bound = 1.25✓ Δ 2.8×10-17 (14 digits) ↗
Equivalence testsEquivalence testslower bound = 2; mode = one_sample_difference; target = 6; upper bound = -2✓ ↗
Equivalence testsEquivalence testslower bound = 2; upper bound = -2✓ ↗
Equivalence testsLarge-offset numerical caselower bound = -2; mode = one_sample_difference; target = 1000000006; upper bound = 2✓ Δ 3.1×10-18 (15+ digits) ↗
Equivalence testsLarge-offset numerical caselower bound = -3; upper bound = 3✓ Δ 5.6×10-17 (15+ digits) ↗
Equivalence testsMissing and excluded populationlower bound = -2; mode = one_sample_difference; target = 6; upper bound = 2✓ Δ 4.3×10-17 (14 digits) ↗
Equivalence testsMissing and excluded populationlower bound = -3; upper bound = 3✓ Δ 8.9×10-16 (14 digits) ↗
Equivalence testsOne-sample difference TOSTlower bound = -2; mode = one_sample_difference; target = 6; upper bound = 2✓ Δ 1.2×10-16 (14 digits) ↗
Equivalence testsWelch two-group TOSTlower bound = -3; upper bound = 3✓ Δ 1.9×10-15 (13 digits) ↗
Factorial ANOVA & ANCOVABartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Factorial ANOVA & ANCOVAChi square associationdefault settings✓ exact ↗
Factorial ANOVA & ANCOVAChi square gofdefault settings✓ exact ↗
Factorial ANOVA & ANCOVAClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAExtreme scaledefault settings✓ Δ 9.4×10-6 (15+ digits) ↗
Factorial ANOVA & ANCOVAExtreme scaletest type = III✓ Δ 9.4×10-6 (15+ digits) ↗
Factorial ANOVA & ANCOVAFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAFactorial ANOVA & ANCOVAconfidence level = 1✓ ↗
Factorial ANOVA & ANCOVAFactorial ANOVA & ANCOVAdefault settings✓ Δ 1.0×10-15 (13 digits) ↗✓ ↗
Factorial ANOVA & ANCOVAFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVAKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Factorial ANOVA & ANCOVALevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVAMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Factorial ANOVA & ANCOVAMissing and excludeddefault settings✓ Δ 5.3×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVAMissing and excludedtest type = III✓ Δ 2.5×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVAMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVANIST AtmWtAg: general lineardefault settings✓ Δ 3.5×10-14 (14 digits) ↗
Factorial ANOVA & ANCOVANIST AtmWtAg: linear effectsdefault settings✓ Δ 2.8×10-14 (14 digits) ↗
Factorial ANOVA & ANCOVANIST Norris: general lineardefault settings✓ Δ 1.2×10-7 (12 digits) ↗
Factorial ANOVA & ANCOVANIST Norris: linear effectsdefault settings✓ Δ 7.2×10-8 (12 digits) ↗
Factorial ANOVA & ANCOVANIST SiRstv: general lineardefault settings✓ Δ 1.0×10-13 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SiRstv: linear effectsdefault settings✓ Δ 1.0×10-13 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs01: general lineardefault settings✓ Δ 1.4×10-14 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs01: linear effectsdefault settings✓ Δ 6.0×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs02: general lineardefault settings✓ Δ 3.7×10-13 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs02: linear effectsdefault settings✓ Δ 2.6×10-13 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs03: general lineardefault settings✓ Δ 2.6×10-11 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs03: linear effectsdefault settings✓ Δ 2.0×10-11 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs04: general lineardefault settings✓ Δ 2.8×10-14 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs04: linear effectsdefault settings✓ Δ 6.2×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs05: general lineardefault settings✓ Δ 3.4×10-13 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs05: linear effectsdefault settings✓ Δ 8.5×10-14 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs06: general lineardefault settings✓ Δ 2.5×10-11 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs06: linear effectsdefault settings✓ Δ 1.5×10-11 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs07: general lineardefault settings✓ Δ 2.8×10-14 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs07: linear effectsdefault settings✓ Δ 6.2×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs08: general lineardefault settings✓ Δ 3.4×10-13 (14 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs08: linear effectsdefault settings✓ Δ 8.5×10-14 (15+ digits) ↗
Factorial ANOVA & ANCOVANIST SmLs09: general lineardefault settings✓ Δ 2.5×10-11 (13 digits) ↗
Factorial ANOVA & ANCOVANIST SmLs09: linear effectsdefault settings✓ Δ 1.5×10-11 (14 digits) ↗
Factorial ANOVA & ANCOVAOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVAOne sample talpha = 0✓ ↗
Factorial ANOVA & ANCOVAOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Factorial ANOVA & ANCOVAOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Factorial ANOVA & ANCOVAOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Factorial ANOVA & ANCOVAOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Factorial ANOVA & ANCOVAOne sample tdefault settings✓ ↗
Factorial ANOVA & ANCOVAOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Factorial ANOVA & ANCOVAOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Factorial ANOVA & ANCOVAOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Factorial ANOVA & ANCOVAPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Factorial ANOVA & ANCOVARegression completioninteractions = group, x; test type = II✓ Δ 2.8×10-14 (14 digits) ↗
Factorial ANOVA & ANCOVARegression completionmax rows = 100✓ Δ 2.2×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingconfidence level = 0.95; max rows = 100✓ Δ 1.4×10-15 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingconfidence level = 0.9; max rows = 100✓ ↗
Factorial ANOVA & ANCOVARegression settingdefault settings✓ Δ 5.8×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingterms = kind interaction, columns x, z✓ Δ 2.8×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingterms = kind main, columns x✓ Δ 2.0×10-16 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingterms = kind nested, columns group, child✓ Δ 2.3×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingterms = kind quadratic, columns x✓ Δ 1.4×10-16 (13 digits) ↗
Factorial ANOVA & ANCOVARegression settingtest type = II✓ Δ 1.4×10-14 (15+ digits) ↗
Factorial ANOVA & ANCOVARegression settingtest type = III✓ Δ 2.5×10-17 (15+ digits) ↗
Factorial ANOVA & ANCOVASTAT REG EFFECTS 001default settings✓ Δ 2.2×10-18 (15+ digits) ↗
Factorial ANOVA & ANCOVASTAT REG GENERAL 001test type = III✓ Δ 2.3×10-18 (15+ digits) ↗
Factorial ANOVA & ANCOVASigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Factorial ANOVA & ANCOVATwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Factorial ANOVA & ANCOVAUnsupporteddefault settings✓ ↗
Factorial ANOVA & ANCOVAWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Factorial ANOVA & ANCOVAWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Factorial ANOVA & ANCOVAWilcoxondefault settings✓ exact ↗
Factorial ANOVA & ANCOVAZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Multivariate ANOVA (MANOVA)Missing and excluded multivariate populationdefault settings✓ Δ 9.9×10-18 (15+ digits) ↗
Multivariate ANOVA (MANOVA)Multivariatedefault settings✓ Δ 6.6×10-16 (12 digits) ↗
Multivariate ANOVA (MANOVA)Multivariate ANOVA (MANOVA)confidence level = 0.2✓ ↗
Multivariate ANOVA (MANOVA)Multivariate ANOVA (MANOVA)default settings✓ Δ 7.3×10-11 (10 digits) ↗✓ ↗
Multivariate ANOVA (MANOVA)Multivariate optiontest type = II✓ Δ 6.6×10-16 (12 digits) ↗
One-sample meanBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
One-sample meanChi square associationdefault settings✓ exact ↗
One-sample meanChi square gofdefault settings✓ exact ↗
One-sample meanClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
One-sample meanFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
One-sample meanFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
One-sample meanKnown sigma Zalpha = 0.1; alternative = greater; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 3.4×10-17 (15+ digits) ↗
One-sample meanKnown sigma Zalpha = 0.1; alternative = less; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 1.3×10-18 (15+ digits) ↗
One-sample meanKnown sigma Zalpha = 0.1; known sigma = 1; max rows = 100; null mean = 0✓ Δ 3.5×10-17 (15+ digits) ↗
One-sample meanKnown sigma Zalpha = 0; known sigma = 2; null mean = 6✓ ↗
One-sample meanKnown sigma Zinput kind = raw; known sigma = 2; null mean = 6✓ Δ 6.7×10-17 (15+ digits) ↗
One-sample meanKnown sigma Zinput kind = summary; known sigma = 2; null mean = 6; sample size = 16; summary mean = 6.8125✓ Δ 6.7×10-17 (15+ digits) ↗
One-sample meanKnown sigma Zknown sigma = 0; null mean = 6✓ ↗
One-sample meanKnown sigma Zknown sigma = 2; null mean = 1000000006✓ Δ 6.7×10-17 (15+ digits) ↗
One-sample meanKnown sigma Zknown sigma = 2; null mean = 6✓ Δ 4.0×10-18 (15+ digits) ↗
One-sample meanKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
One-sample meanLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
One-sample meanMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
One-sample meanMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
One-sample meanOne poisson ratealternative = greater; null rate = 0.15✓ Δ 2.3×10-16 (15+ digits) ↗
One-sample meanOne poisson ratealternative = less; null rate = 0.15✓ Δ 1.1×10-16 (15+ digits) ↗
One-sample meanOne poisson rateexposure unit = hours; input kind = raw; null rate = 0.15✓ Δ 4.4×10-18 (15+ digits) ↗
One-sample meanOne poisson rateinput kind = summary; null rate = 0.15; summary count = 12; summary exposure = 80✓ Δ 4.4×10-18 (15+ digits) ↗
One-sample meanOne poisson ratenull rate = 0.1✓ Δ 7.1×10-18 (15+ digits) ↗
One-sample meanOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
One-sample meanOne sample talpha = 0✓ ↗
One-sample meanOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
One-sample meanOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
One-sample meanOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
One-sample meanOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
One-sample meanOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
One-sample meanOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
One-sample meanOne sample tdefault settings✓ ↗
One-sample meanOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
One-sample meanOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
One-sample meanOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
One-sample meanOne variance chi squarealternative = greater; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
One-sample meanOne variance chi squarealternative = less; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
One-sample meanOne variance chi squareinput kind = raw; null variance = 4; variance display = sd✓ Δ 3.3×10-15 (15+ digits) ↗
One-sample meanOne variance chi squareinput kind = summary; null variance = 4; sample size = 16; summary variance = 3.70917✓ Δ 1.2×10-14 (14 digits) ↗
One-sample meanOne variance chi squarenull variance = 1✓ ↗
One-sample meanOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
One-sample meanPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
One-sample meanPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
One-sample meanSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
One-sample meanTwo poisson ratealternative = greater; input kind = summary; null ratio = 1; summary count = 4; summary count 2 = 8; summary exposure = 38; summary exposure 2 = 42✓ Δ 1.1×10-16 (15+ digits) ↗
One-sample meanTwo poisson ratealternative = greater; null ratio = 1✓ Δ 1.1×10-16 (15+ digits) ↗
One-sample meanTwo poisson ratealternative = less; null ratio = 1✓ Δ 3.6×10-18 (15+ digits) ↗
One-sample meanTwo poisson rateexposure unit = days; input kind = raw; null ratio = 1✓ Δ 5.2×10-17 (15+ digits) ↗
One-sample meanTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
One-sample meanUnsupporteddefault settings✓ ↗
One-sample meanWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
One-sample meanWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
One-sample meanWilcoxondefault settings✓ exact ↗
One-sample meanZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
One-way ANOVAOne-way ANOVA Bonferronialpha = 0.05; procedure = bonferroni✓ Δ 4.6×10-14 (13 digits) ↗
One-way ANOVAOne-way ANOVA Dunnett against control Aalpha = 0.05; control = A; procedure = dunnett✓ Δ 1.9×10-5 (4 digits) ↗
One-way ANOVAOne-way ANOVA Games-Howellalpha = 0.05; procedure = games_howell✓ Δ 7.7×10-10 (7 digits) ↗
One-way ANOVAOne-way ANOVA Helmert contrasts pooledalpha = 0.05; contrast adjustment = none; contrast family = helmert; contrast variance = pooled; procedure = none✓ Δ 4.6×10-14 (13 digits) ↗
One-way ANOVAOne-way ANOVA Holmalpha = 0.05; procedure = holm✓ Δ 4.6×10-14 (13 digits) ↗
One-way ANOVAOne-way ANOVA LSD alpha 0.10alpha = 0.1; procedure = lsd✓ Δ 4.6×10-14 (13 digits) ↗
One-way ANOVAOne-way ANOVA Scheffealpha = 0.05; procedure = scheffe✓ Δ 4.6×10-14 (13 digits) ↗
One-way ANOVAOne-way ANOVA Tukey-Krameralpha = 0.05; procedure = tukey✓ Δ 3.4×10-8 (7 digits) ↗
One-way ANOVAOne-way ANOVA custom contrasts Welch Bonferronialpha = 0.05; coefficients = 1, -0.5, -0.5, 1, 0, -1; contrast adjustment = bonferroni; contrast family = custom; contrast variance = welch; procedure = none✓ Δ 4.6×10-14 (12 digits) ↗
Paired comparisonsAdditional configuration coveragehistogram bins = 5; inference = monte_carlo; max cells = 1000; max rows = 100; mode = paired; resamples = 100; seed = 42✓ Δ 4.3×10-18 (15+ digits) ↗
Paired comparisonsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Paired comparisonsBelow minimum: One response per group is below the independent permutation minimum.inference = exact; mode = independent✓ ↗
Paired comparisonsChi square associationdefault settings✓ exact ↗
Paired comparisonsChi square gofdefault settings✓ exact ↗
Paired comparisonsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Paired comparisonsExact independent mean permutationinference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Paired comparisonsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Paired comparisonsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Paired comparisonsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Paired comparisonsLarge-offset numerical caseinference = exact; mode = independent✓ Δ 3.9×10-17 (15+ digits) ↗
Paired comparisonsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Paired comparisonsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Paired comparisonsMissing and excluded populationinference = exact; mode = independent✓ Δ 8.9×10-16 (14 digits) ↗
Paired comparisonsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Paired comparisonsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Paired comparisonsOne sample talpha = 0✓ ↗
Paired comparisonsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Paired comparisonsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Paired comparisonsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Paired comparisonsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Paired comparisonsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Paired comparisonsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Paired comparisonsOne sample tdefault settings✓ ↗
Paired comparisonsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Paired comparisonsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Paired comparisonsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Paired comparisonsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Paired comparisonsPaired comparisonsalternative = greater; inference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Paired comparisonsPaired comparisonsalternative = less; inference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Paired comparisonsPaired comparisonsinference = exact; max cells = 1000; max rows = 100; mode = independent✓ Δ 1.1×10-16 (15+ digits) ↗
Paired comparisonsPaired comparisonsinference = exact; mode = independent✓ ↗
Paired comparisonsPaired comparisonsinference = exact; mode = paired✓ Δ 4.3×10-18 (15+ digits) ↗
Paired comparisonsPaired comparisonsinference = monte_carlo; mode = independent; resamples = 100; seed = 77✓ Δ 8.8×10-16 (15+ digits) ↗
Paired comparisonsPaired comparisonsinference = unsupported; mode = independent✓ ↗
Paired comparisonsPaired comparisonsmode = independent✓ ↗
Paired comparisonsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Paired comparisonsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Paired comparisonsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Paired comparisonsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Paired comparisonsUnsupporteddefault settings✓ ↗
Paired comparisonsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Paired comparisonsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Paired comparisonsWilcoxondefault settings✓ exact ↗
Paired comparisonsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Rank & median comparisonsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Rank & median comparisonsChi square associationdefault settings✓ exact ↗
Rank & median comparisonsChi square gofdefault settings✓ exact ↗
Rank & median comparisonsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Rank & median comparisonsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Rank & median comparisonsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Rank & median comparisonsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Rank & median comparisonsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Rank & median comparisonsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Rank & median comparisonsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Rank & median comparisonsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Rank & median comparisonsOne sample talpha = 0✓ ↗
Rank & median comparisonsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Rank & median comparisonsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Rank & median comparisonsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Rank & median comparisonsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Rank & median comparisonsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Rank & median comparisonsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Rank & median comparisonsOne sample tdefault settings✓ ↗
Rank & median comparisonsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Rank & median comparisonsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Rank & median comparisonsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Rank & median comparisonsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Rank & median comparisonsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Rank & median comparisonsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Rank & median comparisonsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Rank & median comparisonsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Rank & median comparisonsUnsupporteddefault settings✓ ↗
Rank & median comparisonsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Rank & median comparisonsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Rank & median comparisonsWilcoxondefault settings✓ exact ↗
Rank & median comparisonsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Two independent meansAdditional configuration coveragehistogram bins = 5; inference = monte_carlo; max cells = 1000; max rows = 100; mode = paired; resamples = 100; seed = 42✓ Δ 4.3×10-18 (15+ digits) ↗
Two independent meansBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Two independent meansBelow minimum: One response per group is below the independent permutation minimum.inference = exact; mode = independent✓ ↗
Two independent meansChi square associationdefault settings✓ exact ↗
Two independent meansChi square gofdefault settings✓ exact ↗
Two independent meansClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Two independent meansExact independent mean permutationinference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Two independent meansFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Two independent meansFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Two independent meansKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Two independent meansLarge-offset numerical caseinference = exact; mode = independent✓ Δ 3.9×10-17 (15+ digits) ↗
Two independent meansLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Two independent meansMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Two independent meansMissing and excluded populationinference = exact; mode = independent✓ Δ 8.9×10-16 (14 digits) ↗
Two independent meansMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Two independent meansOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Two independent meansOne sample talpha = 0✓ ↗
Two independent meansOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Two independent meansOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Two independent meansOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Two independent meansOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Two independent meansOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Two independent meansOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Two independent meansOne sample tdefault settings✓ ↗
Two independent meansOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Two independent meansOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Two independent meansOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Two independent meansOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Two independent meansPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Two independent meansPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Two independent meansSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Two independent meansTwo independent meansalternative = greater; inference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Two independent meansTwo independent meansalternative = less; inference = exact; mode = independent✓ Δ 8.8×10-16 (15+ digits) ↗
Two independent meansTwo independent meansinference = exact; max cells = 1000; max rows = 100; mode = independent✓ Δ 1.1×10-16 (15+ digits) ↗
Two independent meansTwo independent meansinference = exact; mode = independent✓ ↗
Two independent meansTwo independent meansinference = exact; mode = paired✓ Δ 4.3×10-18 (15+ digits) ↗
Two independent meansTwo independent meansinference = monte_carlo; mode = independent; resamples = 100; seed = 77✓ Δ 8.8×10-16 (15+ digits) ↗
Two independent meansTwo independent meansinference = unsupported; mode = independent✓ ↗
Two independent meansTwo independent meansmode = independent✓ ↗
Two independent meansTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Two independent meansUnsupporteddefault settings✓ ↗
Two independent meansWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Two independent meansWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Two independent meansWilcoxondefault settings✓ exact ↗
Two independent meansZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Variances & standard deviationsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Variances & standard deviationsChi square associationdefault settings✓ exact ↗
Variances & standard deviationsChi square gofdefault settings✓ exact ↗
Variances & standard deviationsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Variances & standard deviationsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Variances & standard deviationsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zalpha = 0.1; alternative = greater; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 3.4×10-17 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zalpha = 0.1; alternative = less; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 1.3×10-18 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zalpha = 0.1; known sigma = 1; max rows = 100; null mean = 0✓ Δ 3.5×10-17 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zalpha = 0; known sigma = 2; null mean = 6✓ ↗
Variances & standard deviationsKnown sigma Zinput kind = raw; known sigma = 2; null mean = 6✓ Δ 6.7×10-17 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zinput kind = summary; known sigma = 2; null mean = 6; sample size = 16; summary mean = 6.8125✓ Δ 6.7×10-17 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zknown sigma = 0; null mean = 6✓ ↗
Variances & standard deviationsKnown sigma Zknown sigma = 2; null mean = 1000000006✓ Δ 6.7×10-17 (15+ digits) ↗
Variances & standard deviationsKnown sigma Zknown sigma = 2; null mean = 6✓ Δ 4.0×10-18 (15+ digits) ↗
Variances & standard deviationsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Variances & standard deviationsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Variances & standard deviationsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Variances & standard deviationsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Variances & standard deviationsOne poisson ratealternative = greater; null rate = 0.15✓ Δ 2.3×10-16 (15+ digits) ↗
Variances & standard deviationsOne poisson ratealternative = less; null rate = 0.15✓ Δ 1.1×10-16 (15+ digits) ↗
Variances & standard deviationsOne poisson rateexposure unit = hours; input kind = raw; null rate = 0.15✓ Δ 4.4×10-18 (15+ digits) ↗
Variances & standard deviationsOne poisson rateinput kind = summary; null rate = 0.15; summary count = 12; summary exposure = 80✓ Δ 4.4×10-18 (15+ digits) ↗
Variances & standard deviationsOne poisson ratenull rate = 0.1✓ Δ 7.1×10-18 (15+ digits) ↗
Variances & standard deviationsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Variances & standard deviationsOne sample talpha = 0✓ ↗
Variances & standard deviationsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Variances & standard deviationsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Variances & standard deviationsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Variances & standard deviationsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Variances & standard deviationsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Variances & standard deviationsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Variances & standard deviationsOne sample tdefault settings✓ ↗
Variances & standard deviationsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Variances & standard deviationsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Variances & standard deviationsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Variances & standard deviationsOne variance chi squarealternative = greater; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
Variances & standard deviationsOne variance chi squarealternative = less; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
Variances & standard deviationsOne variance chi squareinput kind = raw; null variance = 4; variance display = sd✓ Δ 3.3×10-15 (15+ digits) ↗
Variances & standard deviationsOne variance chi squareinput kind = summary; null variance = 4; sample size = 16; summary variance = 3.70917✓ Δ 1.2×10-14 (14 digits) ↗
Variances & standard deviationsOne variance chi squarenull variance = 1✓ ↗
Variances & standard deviationsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Variances & standard deviationsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Variances & standard deviationsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Variances & standard deviationsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Variances & standard deviationsTwo poisson ratealternative = greater; input kind = summary; null ratio = 1; summary count = 4; summary count 2 = 8; summary exposure = 38; summary exposure 2 = 42✓ Δ 1.1×10-16 (15+ digits) ↗
Variances & standard deviationsTwo poisson ratealternative = greater; null ratio = 1✓ Δ 1.1×10-16 (15+ digits) ↗
Variances & standard deviationsTwo poisson ratealternative = less; null ratio = 1✓ Δ 3.6×10-18 (15+ digits) ↗
Variances & standard deviationsTwo poisson rateexposure unit = days; input kind = raw; null ratio = 1✓ Δ 5.2×10-17 (15+ digits) ↗
Variances & standard deviationsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Variances & standard deviationsUnsupporteddefault settings✓ ↗
Variances & standard deviationsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Variances & standard deviationsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Variances & standard deviationsWilcoxondefault settings✓ exact ↗
Variances & standard deviationsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗

Statistics › Missing data

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Missing-data patterns & diagnosticsEM to a custom tolerance of 1e-10categorical = g; em max iterations = 5000; em preset = custom; em tolerance = 1e-10; quantitative = x, y, z✓ Δ 2.7×10-10 (9 digits) ↗
Missing-data patterns & diagnosticsExcluded rows leave the analysis populationcategorical = g; quantitative = x, y, z✓ Δ 1.7×10-5 (5 digits) ↗
Missing-data patterns & diagnosticsMissing value analysis with the default settingscategorical = g; quantitative = x, y, z✓ Δ 2.7×10-6 (5 digits) ↗
Missing-data patterns & diagnosticsRaw p-values, 1% indicator threshold, SPSS-style EMadjustment = none; categorical = g; em preset = spss; min percent = 1; quantitative = x, y, z✓ Δ 2.6×10-4 (4 digits) ↗
Multiple imputation & pooled regressionNo missing values bypass imputation and give the ordinary OLS fitanalysis = model ols, outcome y, predictors x; confidence level = 0.95; imputations = 2; seed = 20260923; sweeps = 40✓ Δ 1.6×10-15 (15+ digits) ↗
Multiple imputation & pooled regressionPredictive mean matching copies an observed donor valueimputation = 2✓ exact ↗
Multiple imputation & pooled regressionReplaying the completed-dataset operation recreates the recorded tableimputation = 1✓ exact ↗

Statistics › Multivariate & clustering › Clustering

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Hierarchical clusteringConfiguration axiscut k = 3; linkage = ward; mode = rows; standardize = no✓ Δ 4.0×10-19 (15+ digits) ↗
Hierarchical clusteringHierarchical clusteringcut k = 2; linkage = ward✓ Δ 2.2×10-17 (15+ digits) ↗
Hierarchical clusteringHierarchical clusteringcut k = 3; linkage = ward✓ Δ 1.5×10-19 (15+ digits) ↗
Hierarchical clusteringHierarchical clusteringdefault settings✓ ↗
Hierarchical clusteringHierarchical clusteringlinkage = ward; mode = variables✓ ↗
Hierarchical clusteringHierarchical clusteringname = cluster✓ ↗
Hierarchical clusteringHierarchical clusteringname = cluster; unknown option = 1✓ ↗
Hierarchical clusteringHierarchical clusteringname = hcluster✓ exact ↗
Hierarchical clusteringMissing and excluded multivariate populationcut k = 3✓ exact ↗
Hierarchical clusteringMultivariatecut k = 3; linkage = ward✓ exact ↗
Hierarchical clusteringMultivariate optioncut k = 2; linkage = average; mode = variables; standardize = yes✓ Δ 5.5×10-17 (14 digits) ↗
Hierarchical clusteringMultivariate optioncut k = 3; linkage = average✓ Δ 9.7×10-17 (14 digits) ↗
Hierarchical clusteringMultivariate optioncut k = 3; linkage = complete✓ Δ 9.7×10-17 (14 digits) ↗
Hierarchical clusteringOrdered multivariatecut k = 3; linkage = ward✓ exact ↗
Hierarchical clusteringRetained correctionname = hcluster✓ exact ↗
Hierarchical clusteringRetained scoringname = hcluster✓ exact ↗
Hierarchical clusteringVariable-mode average linkagecut k = 2; linkage = average; mode = variables; standardize = yes✓ Δ 5.5×10-17 (14 digits) ↗
K-means clusteringK-means clusteringk = 1✓ ↗
K-means clusteringK-means clusteringk = 2✓ ↗
K-means clusteringK-means clusteringk = 2; restarts = 50; seed = 11✓ Δ 4.9×10-16 (15+ digits) ↗
K-means clusteringK-means clusteringk = 3; restarts = 50; seed = 11✓ Δ 1.1×10-17 (15+ digits) ↗
K-means clusteringK-means clusteringname = cluster✓ ↗
K-means clusteringK-means clusteringname = cluster; unknown option = 1✓ ↗
K-means clusteringK-means clusteringname = kcluster✓ exact ↗
K-means clusteringMissing and excluded multivariate populationk = 3; restarts = 50; seed = 37✓ Δ 2.7×10-17 (15+ digits) ↗
K-means clusteringMultivariatek = 3; restarts = 50; seed = 11✓ Δ 1.9×10-17 (15+ digits) ↗
K-means clusteringMultivariate optionk = 3; restarts = 50; seed = 29; standardize = no✓ Δ 3.3×10-16 (15+ digits) ↗
K-means clusteringOrdered multivariatek = 3; restarts = 50; seed = 11✓ Δ 1.9×10-17 (15+ digits) ↗
K-means clusteringRetained correctionname = kcluster✓ exact ↗
K-means clusteringRetained scoringname = kcluster✓ exact ↗
Model-based clusteringGaussiancovariance = diagonal; k max = 1; k min = 1; kind = gaussian✓ Δ 3.1×10-11 (12 digits) ↗
Model-based clusteringGaussiancovariance = diagonal; k max = 3; k min = 1; kind = gaussian✓ Δ 1.1×10-8 (10 digits) ↗
Model-based clusteringGaussiancovariance = full; k max = 2; k min = 2; kind = gaussian✓ Δ 2.7×10-8 (9 digits) ↗
Model-based clusteringLatent classk max = 1; k min = 1; kind = latent_class✓ Δ 5.0×10-15 (15+ digits) ↗

Statistics › Multivariate & clustering › Dimension reduction

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Canonical correlationCanonical correlation of one variable against onefirst = x1; second = y1✓ Δ 2.3×10-14 (13 digits) ↗
Canonical correlationCanonical correlation of three and two variablesfirst = x1, x2, x3; second = y1, y2✓ Δ 1.4×10-14 (14 digits) ↗
Canonical correlationCanonical correlation of two and two variables with an excluded rowfirst = x1, x3; second = y1, y2✓ Δ 2.4×10-14 (13 digits) ↗
Categorical principal componentsCategorical PCA of numeric variables equals correlation PCAdimensions = 2; numeric = x1, x2, x3✓ Δ 4.3×10-14 (13 digits) ↗
Categorical principal componentsCategorical PCA with a binary nominal variable, one dimensiondimensions = 1; nominal = site; numeric = x1, x2, x3✓ Δ 2.9×10-14 (15+ digits) ↗
Categorical principal componentsCategorical PCA with a binary ordinal variabledimensions = 2; numeric = x1, x2; orders = rating: low < high; ordinal = rating✓ Δ 1.5×10-14 (14 digits) ↗
Correspondence analysisCA principal-coordinate sign conventiondimensions = 1; mode = count_table✓ Δ 1.1×10-16 (15+ digits) ↗
Correspondence analysisCorrespondence analysisdimensions = 0; mode = categorical✓ ↗
Correspondence analysisCorrespondence analysismode = categorical✓ ↗
Correspondence analysisCorrespondence analysismode = count_table✓ Δ 3.4×10-19 (15+ digits) ↗
Correspondence analysisMissing and excluded multivariate populationmode = categorical✓ Δ 4.9×10-16 (15+ digits) ↗
Correspondence analysisMultivariatemode = categorical✓ Δ 2.2×10-16 (15+ digits) ↗
Correspondence analysisMultivariate optionmode = count_table✓ Δ 2.2×10-16 (15+ digits) ↗
Exploratory factor analysisExploratory factor analysisdefault settings✓ ↗
Exploratory factor analysisExploratory factor analysisnames = score✓ ↗
Exploratory factor analysisExploratory factor analysisnames = score; unknown option = 1✓ ↗
Exploratory factor analysisExploratory factor analysisrotation = none✓ Δ 1.1×10-9 (7 digits) ↗
Exploratory factor analysisFactor preprocessing after missing and exclusiondefault settings✓ Δ 5.3×10-15 (14 digits) ↗
Exploratory factor analysisOne-factor maximum-likelihood fitrotation = none✓ Δ 1.0×10-8 (7 digits) ↗
Exploratory factor analysisRetained correctionnames = factor1✓ Δ 1.1×10-16 (15+ digits) ↗
Exploratory factor analysisRetained scoringnames = factor1✓ Δ 4.5×10-11 (10 digits) ↗
Exploratory factor analysisSeeded ML factor fit nonerotation = none✓ Δ 3.2×10-9 (6 digits) ↗
Exploratory factor analysisSeeded ML factor fit varimaxrotation = varimax✓ Δ 4.9×10-10 (7 digits) ↗
Exploratory factor analysisSeeded ML factor-loading sign conventionrotation = none✓ exact ↗
Exploratory factor analysisStable ML factor fit nonerotation = none✓ Δ 3.2×10-9 (6 digits) ↗
Exploratory factor analysisStable ML factor fit varimaxrotation = varimax✓ Δ 4.9×10-10 (7 digits) ↗
Multidimensional scalingClassicaldimensions = 1; source = matrix✓ Δ 1.3×10-15 (15+ digits) ↗
Multidimensional scalingClassicaldimensions = 2; distance = cityblock; labels = unit; objects = rows; source = variables; standardize = no✓ Δ 1.6×10-12 (13 digits) ↗
Multidimensional scalingClassicaldimensions = 2; distance = correlation; labels = unit; objects = variables; source = variables✓ Δ 2.2×10-15 (14 digits) ↗
Multidimensional scalingClassicaldimensions = 2; distance = euclidean; labels = unit; objects = rows; source = variables; standardize = yes✓ Δ 3.9×10-14 (14 digits) ↗
Multidimensional scalingClassicaldimensions = 2; labels = site; source = matrix✓ Δ 2.8×10-14 (14 digits) ↗
Multidimensional scalingOrdinaldimensions = 2; source = matrix; ties = primary✓ Δ 1.3×10-14 (14 digits) ↗
Multidimensional scalingOrdinaldimensions = 2; source = matrix; ties = secondary✓ Δ 1.3×10-14 (14 digits) ↗
Multiple correspondence analysisIndicator MCA of three variables, complete rowsdimensions = 2; missing = complete✓ Δ 2.3×10-13 (13 digits) ↗
Multiple correspondence analysisIndicator MCA reporting one dimensiondimensions = 1; missing = complete✓ Δ 2.3×10-13 (13 digits) ↗
Multiple correspondence analysisIndicator MCA with a supplementary variabledimensions = 1; missing = complete; supplementary = sector✓ Δ 2.3×10-13 (13 digits) ↗
Multiple correspondence analysisIndicator MCA with missing values as a categorydimensions = 2; missing = category✓ Δ 4.2×10-14 (13 digits) ↗
Principal component analysisMissing and excluded multivariate populationn components = 2✓ Δ 2.7×10-16 (15+ digits) ↗
Principal component analysisMultivariaten components = 2; scaling = correlation✓ Δ 6.7×10-16 (15+ digits) ↗
Principal component analysisMultivariate optionn components = 2; scaling = covariance✓ Δ 1.4×10-16 (15+ digits) ↗
Principal component analysisPrincipal component analysisdefault settings✓ ↗
Principal component analysisPrincipal component analysisn components = 0✓ ↗
Principal component analysisPrincipal component analysisn components = 1; scaling = correlation✓ Δ 4.4×10-17 (15+ digits) ↗
Principal component analysisPrincipal component analysisn components = 2; scaling = correlation✓ Δ 2.6×10-16 (15+ digits) ↗
Principal component analysisPrincipal component analysisnames = pc1✓ Δ 5.6×10-16 (15+ digits) ↗
Principal component analysisPrincipal component analysisnames = score✓ ↗
Principal component analysisPrincipal component analysisnames = score; unknown option = 1✓ ↗
Principal component analysisRetained correctionnames = pc1✓ Δ 1.8×10-15 (14 digits) ↗
Principal component analysisRetained scoringnames = pc1✓ Δ 6.9×10-11 (10 digits) ↗

Statistics › Proportions & categorical data

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Categorical goodness-of-fitBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Categorical goodness-of-fitChi square associationdefault settings✓ exact ↗
Categorical goodness-of-fitChi square gofdefault settings✓ exact ↗
Categorical goodness-of-fitClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Categorical goodness-of-fitFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Categorical goodness-of-fitFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Categorical goodness-of-fitKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Categorical goodness-of-fitLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Categorical goodness-of-fitMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Categorical goodness-of-fitMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Categorical goodness-of-fitOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Categorical goodness-of-fitOne sample talpha = 0✓ ↗
Categorical goodness-of-fitOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Categorical goodness-of-fitOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Categorical goodness-of-fitOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Categorical goodness-of-fitOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Categorical goodness-of-fitOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Categorical goodness-of-fitOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Categorical goodness-of-fitOne sample tdefault settings✓ ↗
Categorical goodness-of-fitOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Categorical goodness-of-fitOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Categorical goodness-of-fitOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Categorical goodness-of-fitOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Categorical goodness-of-fitPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Categorical goodness-of-fitPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Categorical goodness-of-fitSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Categorical goodness-of-fitTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Categorical goodness-of-fitUnsupporteddefault settings✓ ↗
Categorical goodness-of-fitWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Categorical goodness-of-fitWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Categorical goodness-of-fitWilcoxondefault settings✓ exact ↗
Categorical goodness-of-fitZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Contingency tables & association2 x 2 crosstab at 90%: signed phi, odds ratio and relative risk intervalsconfidence = 0.9; haldane correction = no; row = treated✓ Δ 8.4×10-16 (15+ digits) ↗
Contingency tables & association2 x 2 crosstab with an empty cell and no correctionconfidence = 0.95; haldane correction = no; row = treated✓ Δ 4.4×10-18 (15+ digits) ↗
Contingency tables & association2 x 2 crosstab with an empty cell and the 0.5 correctionconfidence = 0.95; haldane correction = yes; row = treated✓ Δ 9.0×10-16 (15+ digits) ↗
Contingency tables & association3 x 3 crosstab: chi-square family, nominal, ordinal, linear-by-linear and etaconfidence = 0.95; haldane correction = no; row = exposure; row order = Low, Medium, High✓ Δ 5.5×10-15 (14 digits) ↗
Contingency tables & associationCategoricalinclude margins = no; kind = categorical✓ exact ↗
Contingency tables & associationCategoricalinclude margins = yes; kind = categorical; overall percentages = yes; row percentages = yes✓ exact ↗
Contingency tables & associationNumerickind = numeric✓ Δ 1.1×10-9 (8 digits) ↗✓ Δ 1.2×10-7 (15+ digits) ↗
Loglinear modelsSTAT REG LOGLINEAR GENERAL 001coding = last; conditional test = off; confidence level = 0.95; count = n; sampling = poisson; terms = A*B, C✓ Δ 6.3×10-10 (8 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR GENERAL 002coding = first; conditional test = off; confidence level = 0.9; count = n; sampling = multinomial; terms = A*B, C✓ Δ 1.3×10-9 (8 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR GENERAL 003coding = last; conditional test = off; confidence level = 0.95; count = n; fixed margins = A; sampling = product_multinomial; terms = B*C✓ Δ 1.4×10-8 (7 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR GENERAL 004cell structure = weight; coding = last; conditional test = off; confidence level = 0.95; count = n; intervals = wald_and_profile; pseudocount = yes; reference levels = B=b2; sampling = poisson; terms = A*B, C✓ Δ 3.4×10-13 (12 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR GENERAL 005coding = last; conditional test = off; confidence level = 0.95; count = n; sampling = poisson; structural zeros = A=a1; B=b3; C=c2; terms = A*B, C✓ Δ 9.7×10-12 (10 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR LOGIT 001coding = last; conditional test = off; confidence level = 0.95; count = n; dependent = C; sampling = poisson; terms = A*C, B*C✓ Δ 4.9×10-11 (9 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR LOGIT 002coding = first; conditional test = off; confidence level = 0.95; count = n; dependent = C; sampling = multinomial; terms = A*C✓ Δ 9.2×10-13 (11 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR SELECTION 001alpha = 0.05; coding = last; conditional test = off; confidence level = 0.95; count = n; highest order only = no; sampling = poisson✓ Δ 8.8×10-15 (12 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR SELECTION 002alpha = 0.3; coding = last; conditional test = off; confidence level = 0.95; count = n; highest order only = no; sampling = poisson✓ Δ 8.8×10-15 (12 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR SELECTION 003alpha = 0.05; coding = last; conditional test = off; confidence level = 0.95; count = n; highest order only = yes; sampling = poisson✓ Δ 8.8×10-15 (12 digits) ↗
Loglinear modelsSTAT REG LOGLINEAR SELECTION 004alpha = 0.05; coding = last; conditional test = off; confidence level = 0.95; count = n; fixed margins = A, B; highest order only = no; sampling = product_multinomial✓ Δ 8.8×10-15 (12 digits) ↗
Multiple-response tablesMultiple-category frequencies: repeated codes count as responses once per caseset name = choices✓ Δ 3.0×10-16 (15+ digits) ↗
Multiple-response tablesMultiple-dichotomy frequencies: case base, responses and percentsset name = features✓ Δ 3.5×10-16 (15+ digits) ↗
Multiple-response tablesMultiple-response crosstab against a column on the cases basebase = cases; set name = features✓ Δ 4.0×10-16 (15+ digits) ↗
Multiple-response tablesMultiple-response crosstab of two sets on the responses basebase = responses; by set = name choices, kind category, columns c1, c2, counted value None, label Suppliers named; by set name = choices; set name = features✓ Δ 3.7×10-16 (15+ digits) ↗
One proportionBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
One proportionChi square associationdefault settings✓ exact ↗
One proportionChi square gofdefault settings✓ exact ↗
One proportionClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
One proportionFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
One proportionFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
One proportionKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
One proportionLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
One proportionMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
One proportionMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
One proportionOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
One proportionOne sample talpha = 0✓ ↗
One proportionOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
One proportionOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
One proportionOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
One proportionOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
One proportionOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
One proportionOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
One proportionOne sample tdefault settings✓ ↗
One proportionOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
One proportionOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
One proportionOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
One proportionOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
One proportionPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
One proportionPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
One proportionSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
One proportionTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
One proportionUnsupporteddefault settings✓ ↗
One proportionWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
One proportionWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
One proportionWilcoxondefault settings✓ exact ↗
One proportionZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zalpha = 0.1; alternative = greater; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 3.4×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zalpha = 0.1; alternative = less; confidence level = 0.9; known sigma = 2; max rows = 100; null mean = 6✓ Δ 1.3×10-18 (15+ digits) ↗
Poisson ratesKnown sigma Zalpha = 0.1; known sigma = 1; max rows = 100; null mean = 0✓ Δ 3.5×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zalpha = 0; known sigma = 2; null mean = 6✓ ↗
Poisson ratesKnown sigma Zinput kind = raw; known sigma = 2; null mean = 6✓ Δ 6.7×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zinput kind = summary; known sigma = 2; null mean = 6; sample size = 16; summary mean = 6.8125✓ Δ 6.7×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zknown sigma = 0; null mean = 6✓ ↗
Poisson ratesKnown sigma Zknown sigma = 2; null mean = 1000000006✓ Δ 6.7×10-17 (15+ digits) ↗
Poisson ratesKnown sigma Zknown sigma = 2; null mean = 6✓ Δ 4.0×10-18 (15+ digits) ↗
Poisson ratesOne poisson ratealternative = greater; null rate = 0.15✓ Δ 2.3×10-16 (15+ digits) ↗
Poisson ratesOne poisson ratealternative = less; null rate = 0.15✓ Δ 1.1×10-16 (15+ digits) ↗
Poisson ratesOne poisson rateexposure unit = hours; input kind = raw; null rate = 0.15✓ Δ 4.4×10-18 (15+ digits) ↗
Poisson ratesOne poisson rateinput kind = summary; null rate = 0.15; summary count = 12; summary exposure = 80✓ Δ 4.4×10-18 (15+ digits) ↗
Poisson ratesOne poisson ratenull rate = 0.1✓ Δ 7.1×10-18 (15+ digits) ↗
Poisson ratesOne variance chi squarealternative = greater; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
Poisson ratesOne variance chi squarealternative = less; null variance = 4✓ Δ 3.3×10-15 (15+ digits) ↗
Poisson ratesOne variance chi squareinput kind = raw; null variance = 4; variance display = sd✓ Δ 3.3×10-15 (15+ digits) ↗
Poisson ratesOne variance chi squareinput kind = summary; null variance = 4; sample size = 16; summary variance = 3.70917✓ Δ 1.2×10-14 (14 digits) ↗
Poisson ratesOne variance chi squarenull variance = 1✓ ↗
Poisson ratesTwo poisson ratealternative = greater; input kind = summary; null ratio = 1; summary count = 4; summary count 2 = 8; summary exposure = 38; summary exposure 2 = 42✓ Δ 1.1×10-16 (15+ digits) ↗
Poisson ratesTwo poisson ratealternative = greater; null ratio = 1✓ Δ 1.1×10-16 (15+ digits) ↗
Poisson ratesTwo poisson ratealternative = less; null ratio = 1✓ Δ 3.6×10-18 (15+ digits) ↗
Poisson ratesTwo poisson rateexposure unit = days; input kind = raw; null ratio = 1✓ Δ 5.2×10-17 (15+ digits) ↗
Related categorical outcomesBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Related categorical outcomesChi square associationdefault settings✓ exact ↗
Related categorical outcomesChi square gofdefault settings✓ exact ↗
Related categorical outcomesClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Related categorical outcomesFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Related categorical outcomesFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Related categorical outcomesKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Related categorical outcomesLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Related categorical outcomesMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Related categorical outcomesMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Related categorical outcomesOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Related categorical outcomesOne sample talpha = 0✓ ↗
Related categorical outcomesOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Related categorical outcomesOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Related categorical outcomesOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Related categorical outcomesOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Related categorical outcomesOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Related categorical outcomesOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Related categorical outcomesOne sample tdefault settings✓ ↗
Related categorical outcomesOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Related categorical outcomesOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Related categorical outcomesOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Related categorical outcomesOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Related categorical outcomesPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Related categorical outcomesPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Related categorical outcomesSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Related categorical outcomesTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Related categorical outcomesUnsupporteddefault settings✓ ↗
Related categorical outcomesWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Related categorical outcomesWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Related categorical outcomesWilcoxondefault settings✓ exact ↗
Related categorical outcomesZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Stratified categorical associationBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Stratified categorical associationChi square associationdefault settings✓ exact ↗
Stratified categorical associationChi square gofdefault settings✓ exact ↗
Stratified categorical associationClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Stratified categorical associationFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Stratified categorical associationFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Stratified categorical associationKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Stratified categorical associationLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Stratified categorical associationMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Stratified categorical associationMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Stratified categorical associationOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Stratified categorical associationOne sample talpha = 0✓ ↗
Stratified categorical associationOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Stratified categorical associationOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Stratified categorical associationOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Stratified categorical associationOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Stratified categorical associationOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Stratified categorical associationOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Stratified categorical associationOne sample tdefault settings✓ ↗
Stratified categorical associationOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Stratified categorical associationOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Stratified categorical associationOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Stratified categorical associationOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Stratified categorical associationPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Stratified categorical associationPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Stratified categorical associationSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Stratified categorical associationTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Stratified categorical associationUnsupporteddefault settings✓ ↗
Stratified categorical associationWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Stratified categorical associationWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Stratified categorical associationWilcoxondefault settings✓ exact ↗
Stratified categorical associationZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Trend in proportionsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Trend in proportionsChi square associationdefault settings✓ exact ↗
Trend in proportionsChi square gofdefault settings✓ exact ↗
Trend in proportionsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Trend in proportionsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Trend in proportionsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Trend in proportionsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Trend in proportionsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Trend in proportionsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Trend in proportionsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Trend in proportionsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Trend in proportionsOne sample talpha = 0✓ ↗
Trend in proportionsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Trend in proportionsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Trend in proportionsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Trend in proportionsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Trend in proportionsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Trend in proportionsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Trend in proportionsOne sample tdefault settings✓ ↗
Trend in proportionsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Trend in proportionsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Trend in proportionsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Trend in proportionsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Trend in proportionsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Trend in proportionsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Trend in proportionsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Trend in proportionsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Trend in proportionsUnsupporteddefault settings✓ ↗
Trend in proportionsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Trend in proportionsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Trend in proportionsWilcoxondefault settings✓ exact ↗
Trend in proportionsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Two independent proportionsBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Two independent proportionsChi square associationdefault settings✓ exact ↗
Two independent proportionsChi square gofdefault settings✓ exact ↗
Two independent proportionsClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Two independent proportionsFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Two independent proportionsFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Two independent proportionsKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Two independent proportionsLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Two independent proportionsMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Two independent proportionsMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Two independent proportionsOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Two independent proportionsOne sample talpha = 0✓ ↗
Two independent proportionsOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Two independent proportionsOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Two independent proportionsOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Two independent proportionsOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Two independent proportionsOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Two independent proportionsOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Two independent proportionsOne sample tdefault settings✓ ↗
Two independent proportionsOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Two independent proportionsOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Two independent proportionsOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Two independent proportionsOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Two independent proportionsPaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Two independent proportionsPooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Two independent proportionsSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Two independent proportionsTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Two independent proportionsUnsupporteddefault settings✓ ↗
Two independent proportionsWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Two independent proportionsWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Two independent proportionsWilcoxondefault settings✓ exact ↗
Two independent proportionsZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗

Statistics › Repeated measures & mixed models

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Binomial and Poisson mixed modelsBinomialevent value = 1; family = binomial✓ Δ 7.1×10-15 (15+ digits) ↗
Binomial and Poisson mixed modelsPoissonfamily = poisson✓ Δ 7.4×10-15 (15+ digits) ↗
Generalized estimating equationsBinomialfamily = binomial; working correlation = exchangeable✓ Δ 4.5×10-16 (14 digits) ↗
Generalized estimating equationsPoissonfamily = poisson; working correlation = independence✓ Δ 3.3×10-16 (14 digits) ↗
Linear mixed modelsLarge-offset random-slope mixed modelestimation = reml; model mode = random_slope✓ Δ 3.5×10-14 (13 digits) ↗
Linear mixed modelsLinear mixed modelsbad setting = 1; estimation = reml✓ ↗
Linear mixed modelsLinear mixed modelsbad setting = 1; estimation = reml; model mode = random_slope✓ ↗
Linear mixed modelsLinear mixed modelscomponents = name a, level column a, relation crossed, name ab, level column ab, relation nested, parent column a; estimation = reml; max rows = 1000; model mode = variance_components✓ Δ 4.8×10-12 (11 digits) ↗
Linear mixed modelsLinear mixed modelscomponents = name a, level column a, relation crossed, name b, level column b, relation crossed; estimation = ml; max rows = 1000; model mode = variance_components✓ Δ 2.9×10-13 (12 digits) ↗
Linear mixed modelsLinear mixed modelscomponents = name a, level column a, relation crossed, name b, level column b, relation crossed; estimation = reml; max rows = 1000; model mode = variance_components✓ Δ 2.5×10-11 (10 digits) ↗
Linear mixed modelsLinear mixed modelsestimation = ml✓ Δ 4.4×10-16 (15+ digits) ↗
Linear mixed modelsLinear mixed modelsestimation = reml✓ Δ 8.1×10-11 (9 digits) ↗✓ ↗
Linear mixed modelsLinear mixed modelsestimation = reml; model mode = random_slope✓ Δ 4.4×10-8 (7 digits) ↗✓ ↗
Linear mixed modelsRegression edgeestimation = reml✓ Δ 4.1×10-11 (9 digits) ↗
Linear mixed modelsRegression edgeestimation = reml; model mode = random_slope✓ ↗
Panel fixed-effects regressionPanel fixed-effects regressionbad setting = 1✓ ↗
Panel fixed-effects regressionPanel fixed-effects regressionconfidence level = 0.9✓ Δ 2.8×10-18 (15+ digits) ↗
Panel fixed-effects regressionPanel fixed-effects regressiondefault settings✓ Δ 4.4×10-16 (15+ digits) ↗✓ ↗
Panel fixed-effects regressionRegression edgedefault settings✓ Δ 4.4×10-16 (15+ digits) ↗
Rank comparisons & concordanceBartlettdefault settings✓ Δ 1.5×10-18 (15+ digits) ↗
Rank comparisons & concordanceChi square associationdefault settings✓ exact ↗
Rank comparisons & concordanceChi square gofdefault settings✓ exact ↗
Rank comparisons & concordanceClassicposthoc = no✓ Δ 3.2×10-15 (14 digits) ↗
Rank comparisons & concordanceFdefault settings✓ Δ 6.2×10-15 (14 digits) ↗
Rank comparisons & concordanceFishersuccess value = yes✓ Δ 2.2×10-16 (15+ digits) ↗
Rank comparisons & concordanceFriedmanalternative = less✓ ↗
Rank comparisons & concordanceFriedmandefault settings✓ Δ 3.1×10-18 (15+ digits) ↗✓ ↗
Rank comparisons & concordanceFriedmanmax blocks = 10; max rows = 100✓ Δ 2.3×10-18 (15+ digits) ↗
Rank comparisons & concordanceKruskal wallisdefault settings✓ Δ 4.4×10-15 (13 digits) ↗
Rank comparisons & concordanceLevenedefault settings✓ Δ 1.1×10-16 (15+ digits) ↗
Rank comparisons & concordanceMann whitneydefault settings✓ Δ 3.1×10-18 (15+ digits) ↗
Rank comparisons & concordanceMoods mediandefault settings✓ Δ 3.0×10-15 (14 digits) ↗
Rank comparisons & concordanceOne proportionsuccess value = yes; target = 0.4✓ Δ 2.8×10-16 (15+ digits) ↗
Rank comparisons & concordanceOne sample talpha = 0✓ ↗
Rank comparisons & concordanceOne sample talpha = 0.1; target = 0✓ Δ 3.9×10-18 (15+ digits) ↗
Rank comparisons & concordanceOne sample talpha = 0.1; target = 5.96✓ Δ 2.0×10-15 (13 digits) ↗
Rank comparisons & concordanceOne sample talpha = 0.1; target = 5.97688✓ Δ 3.0×10-16 (14 digits) ↗
Rank comparisons & concordanceOne sample talternative = greater; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Rank comparisons & concordanceOne sample talternative = less; target = 6.2✓ Δ 1.6×10-15 (14 digits) ↗
Rank comparisons & concordanceOne sample talternative = two-sided; target = 6.2✓ Δ 1.6×10-15 (14 digits) ⟨R⟩ ↗
Rank comparisons & concordanceOne sample tdefault settings✓ ↗
Rank comparisons & concordanceOne sample ttarget = 0✓ Δ 4.8×10-16 (15+ digits) ↗✓ ↗
Rank comparisons & concordanceOne sample ttarget = 1000000006✓ Δ 1.1×10-8 (7 digits) ↗
Rank comparisons & concordanceOne sample ttarget = 6✓ Δ 2.6×10-15 (14 digits) ↗
Rank comparisons & concordanceOne way ANOVAposthoc = yes✓ Δ 7.1×10-15 (14 digits) ↗
Rank comparisons & concordancePaired tdefault settings✓ Δ 1.5×10-14 (13 digits) ⟨R⟩ ↗
Rank comparisons & concordancePooleddefault settings✓ Δ 8.9×10-16 (14 digits) ↗
Rank comparisons & concordanceRunsalternative = greater; inference = exact; reference mode = fixed; reference value = 6✓ Δ 1.4×10-18 (15+ digits) ↗
Rank comparisons & concordanceRunsalternative = less; inference = exact; reference mode = fixed; reference value = 6✓ Δ 1.0×10-19 (15+ digits) ↗
Rank comparisons & concordanceRunsalternative = two-sided; inference = asymptotic; max exact n = 20; reference mode = median✓ Δ 1.7×10-18 (15+ digits) ↗
Rank comparisons & concordanceRunsdefault settings✓ ↗
Rank comparisons & concordanceRunsinference = asymptotic✓ Δ 1.7×10-18 (15+ digits) ↗
Rank comparisons & concordanceRunsinference = asymptotic; reference mode = median✓ ↗
Rank comparisons & concordanceSigndefault settings✓ Δ 6.4×10-21 (15+ digits) ↗
Rank comparisons & concordanceTwo way ANOVAposthoc = no✓ Δ 5.7×10-15 (12 digits) ↗
Rank comparisons & concordanceUnsupporteddefault settings✓ ↗
Rank comparisons & concordanceWelchdefault settings✓ Δ 6.6×10-15 (15+ digits) ⟨R⟩ ↗
Rank comparisons & concordanceWelchposthoc = no✓ Δ 2.6×10-15 (14 digits) ↗
Rank comparisons & concordanceWilcoxondefault settings✓ exact ↗
Rank comparisons & concordanceZsuccess value = yes✓ Δ 5.8×10-17 (15+ digits) ↗
Repeated-measures ANOVARepeated measures one within factor three levelscells = t1, t2, t3; contrast family = none; pairwise procedure = none✓ Δ 4.5×10-9 (9 digits) ↗
Repeated-measures ANOVARepeated measures two occasions equals paired t squaredcells = t1, t2; contrast family = none; pairwise procedure = none✓ Δ 9.7×10-15 (14 digits) ↗
Repeated-measures ANOVARepeated measures two within factors 2 x 2cells = t1, t2, t3, t4; contrast family = none; pairwise procedure = none✓ Δ 7.1×10-14 (13 digits) ↗
Repeated-measures ANOVARepeated measures with a balanced between-subject factorbetween = grp; cells = t1, t2, t3; contrast family = none; pairwise procedure = none✓ Δ 6.9×10-9 (10 digits) ↗
Repeated-measures covariance modelsBalanced random-intercept Satterthwaite df equal the classical stratum dfconfidence level = 0.95; estimation = reml; max rows = 20000; occasion spacing = None; random intercept = yes; residual structure = independent✓ Δ 6.4×10-14 (14 digits) ↗
Repeated-measures covariance modelsIndependent equal-variance REML repeated-measures model reproduces OLSconfidence level = 0.95; estimation = reml; max rows = 20000; occasion spacing = None; random intercept = no; residual structure = independent✓ Δ 4.3×10-8 (7 digits) ↗

Statistics › Summaries & tables

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Case listing & summariesCase summaries by group: five listed, statistics on every caselimit = 5; metadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; statistics = yes✓ Δ 2.8×10-16 (15+ digits) ↗
Case listing & summariesCase summaries under an integer frequency weightlimit = 100; metadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; statistics = yes✓ Δ 3.4×10-15 (15+ digits) ↗
Custom table builderCustom table: column proportion comparisons with Holm lettersadjustment = holm; alpha = 0.05; compare = proportions; independence = no; rows = dept; statistics = count, column_percent; totals = yes✓ Δ 1.1×10-16 (14 digits) ↗
Custom table builderCustom table: counts, column, row and table percents, chi-squareadjustment = holm; alpha = 0.05; compare = none; independence = yes; rows = dept; statistics = count, column_percent, row_percent, table_percent; totals = yes✓ Δ 4.0×10-16 (15+ digits) ↗
Custom table builderCustom table: nested rows with subtotals and outer-group column percentsadjustment = holm; alpha = 0.05; compare = none; independence = no; rows = dept, level; statistics = count, column_percent, layer_percent; totals = yes✓ Δ 4.0×10-16 (15+ digits) ↗
Custom table builderCustom table: scale summaries and Welch mean comparisons with Bonferroniadjustment = bonferroni; alpha = 0.05; compare = means; independence = no; rows = dept; statistics = mean, sd, median, valid_n, count; summary = sales; totals = yes✓ Δ 4.6×10-16 (14 digits) ↗
Descriptive statisticsCategoricalinclude margins = no; kind = categorical✓ exact ↗
Descriptive statisticsCategoricalinclude margins = yes; kind = categorical; overall percentages = yes; row percentages = yes✓ exact ↗
Descriptive statisticsDescriptive summary of a column with user-missing values maskedmetadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...✓ Δ 1.6×10-16 (15+ digits) ↗
Descriptive statisticsDescriptive summary under a frequency weight after exclusionsmetadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...✓ Δ 4.2×10-16 (15+ digits) ↗
Descriptive statisticsExplore by factor with Tukey's hinges, a 90% interval and exclusionsconfidence level = 0.9; dependents = score; metadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; quartiles = tukey_hinges✓ Δ 8.8×10-9 (9 digits) ↗
Descriptive statisticsExplore one dependent: descriptives, M-estimators, extremes, normality, boxconfidence level = 0.95; dependents = score; metadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; quartiles = haverage✓ Δ 3.9×10-9 (9 digits) ↗
Descriptive statisticsNumerickind = numeric✓ Δ 1.1×10-9 (8 digits) ↗✓ Δ 1.2×10-7 (15+ digits) ↗
FrequenciesFrequencies cut points after row exclusions, no statisticscut points = 4; metadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; quartiles = no; statistics = no; user missing valid = no✓ Δ 3.4×10-15 (15+ digits) ↗
FrequenciesFrequencies treating user-missing values as validmetadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; percentiles = 50; quartiles = no; statistics = yes; user missing valid = yes✓ Δ 5.8×10-14 (15+ digits) ↗
FrequenciesFrequencies with labels, user-missing codes and type-6 percentilesmetadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...; percentiles = 10, 90; quartiles = yes; statistics = yes; user missing valid = no✓ Δ 1.1×10-14 (15+ digits) ↗
Variable codebookCodebook: scale summary, labelled categories and missing countsmetadata = score variable label Test score, user missing values 999, grade variable label Grade, measurement level nominal, valu...✓ Δ 4.0×10-15 (15+ digits) ↗

Statistics › Survival

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Actuarial life tablesActuarial life table with 2-unit intervals to 10censored value = 0; event value = 1; interval width = 2; last boundary = 10; status = status; time = time✓ Δ 1.0×10-17 (15+ digits) ↗
Actuarial life tablesScheduled visits with text status values and the grouped PH score testboundaries = 0, 1, 2, 3, 4, 5, 6; censored value = censored; event value = event; observation model = scheduled_visits; status = status; strata = group; time = time✓ Δ 1.9×10-17 (15+ digits) ↗
Actuarial life tablesUnequal boundaries with two strata and the grouped Gehan testboundaries = 0, 1, 3, 6, 12; censored value = 0; event value = 1; status = status; strata = group; time = time✓ Δ 2.9×10-15 (14 digits) ↗
Cox regressionBreslow ties with the robust sandwich variancecategorical covariates = trt; censored value = 0; event value = 1; layout = counting_process; numeric covariates = x; robust = yes; start = start; status = status; stop = stop; subject = id; ties = breslow✓ Δ 7.4×10-15 (14 digits) ↗
Cox regressionCounting-process Cox model with Efron tiescategorical covariates = trt; censored value = 0; event value = 1; layout = counting_process; numeric covariates = x; start = start; status = status; stop = stop; subject = id✓ Δ 1.4×10-14 (14 digits) ↗
Cox regressionCox regressioncensored value = 0; confidence level = 0.9; event value = 1✓ ↗
Cox regressionCox regressioncensored value = 0; event value = 1✓ Δ 8.7×10-10 (8 digits) ↗✓ ↗
Cox regressionOne row per subject with the covariate-by-time term x tcategorical covariates = trt; censored value = 0; event value = 1; layout = one_row; numeric covariates = x; status = status; stop = time; time covariates = x; time function = time✓ Δ 1.6×10-10 (11 digits) ↗
Cox regressionOne row per subject with the step function x 1{t > 8}categorical covariates = trt; censored value = 0; event value = 1; layout = one_row; numeric covariates = x; status = status; step time = 8; stop = time; time covariates = x; time function = step✓ Δ 3.1×10-11 (11 digits) ↗
Cox regressionStratified baseline hazards by sex with 90% intervalscategorical covariates = trt; censored value = 0; confidence level = 0.9; event value = 1; layout = counting_process; numeric covariates = x; start = start; status = status; stop = stop; strata = sex; subject = id✓ Δ 1.1×10-13 (13 digits) ↗
Cox regressionSurvivalcensored value = 0; event value = 1✓ Δ 3.7×10-18 (15+ digits) ↗
Cox regressionSurvival corrected optioncensored value = 0; confidence level = 0.95; event value = 1; max diagnostic rows = 100; max iterations = 300✓ Δ 8.6×10-14 (11 digits) ↗
Cox regressionSurvival optioncensored value = 0; confidence level = 0.95; event value = 1; max diagnostic rows = 100; max iterations = 300✓ Δ 4.7×10-14 (13 digits) ↗
Cox regressionSurvival optioncensored value = 0; confidence level = 0.9; event value = 1; max diagnostic rows = 100; max iterations = 300✓ ↗
Cox regressionSurvival populationcensored value = 0; event value = 1✓ Δ 1.9×10-14 (13 digits) ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; confidence level = 0.4; event value = 1✓ ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; event value = 1✓ Δ 2.5×10-11 (15+ digits) ↗✓ ↗not assessed
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; event value = 1; times = 0, 1, 2✓ exact ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; event value = 1; times = 1, 5, 6✓ exact ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; event value = 1; times = 1000137, 1001370, 1005500✓ exact ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 0; event value = 1; times = six✓ ↗
Kaplan-Meier survival curvesKaplan-Meier survival curvescensored value = 1; event value = 1✓ ↗
Kaplan-Meier survival curvesSurvivalcensored value = 0; event value = 1✓ Δ 4.5×10-16 (15+ digits) ↗
Kaplan-Meier survival curvesSurvivalcensored value = 0; event value = 1; times = 6, 7, 10, 23✓ exact ↗
Kaplan-Meier survival curvesSurvival optioncensored value = 0; confidence level = 0.9; event value = 1✓ Δ 1.1×10-16 (15+ digits) ↗
Kaplan-Meier survival curvesSurvival populationcensored value = 0; event value = 1✓ exact ↗
Kaplan-Meier survival curvesSurvival populationcensored value = 0; event value = 1; times = 7, 10, 15, 20✓ exact ↗

Survey

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Calibrate weightsno comparison case yet
Count regressionno comparison case yet
Cox regressionSurvey Cox regression (Taylor)confidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; event value = event; status = status; time = time✓ Δ 1.4×10-14 (14 digits) ↗
Cox regressionSurvey Cox regression from supplied replicate weights with 90% intervalsconfidence level = 0.9; covariates = x; design = weight w, variance replicate, replicates columns rw1, rw2, rw3, rw4, method other, weight type full, scale 0.75, rsca...; event value = event; status = status; time = time✓ Δ 5.0×10-9 (8 digits) ↗
Frequencies and crosstabsSurvey crosstab with a first-stage FPC and 90% intervalsconfidence level = 0.9; design = weight w, variance linearization, stage1 strata h, units psu, fpc N, fpc kind count✓ Δ 1.4×10-14 (12 digits) ↗
Frequencies and crosstabsSurvey crosstab with the second-order Rao-Scott Fconfidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu✓ Δ 6.4×10-14 (12 digits) ↗
Frequencies and crosstabsSurvey one-way frequencies in a domainconfidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu; domain values = B✓ Δ 2.3×10-13 (13 digits) ↗
Generate replicate weightsno comparison case yet
Item consistency (Cronbach's alpha)Configuration axisaggregation = sum; names = scale_sum✓ exact ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)aggregation = mean; names = scale_score✓ exact ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)aggregation = mean; names = score✓ ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)aggregation = mean; names = score; unknown option = 1✓ ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)items = i1✓ ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)items = i1, i2✓ exact ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)items = i1, i2, i3✓ Δ 1.9×10-18 (15+ digits) ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)items = i1, i2; reverse items = i3; scale high = 5; scale low = 1✓ ↗
Item consistency (Cronbach's alpha)Item consistency (Cronbach's alpha)items = i1, i3✓ exact ↗
Item consistency (Cronbach's alpha)Missing and excluded multivariate populationitems = i1, i2, i3✓ Δ 2.2×10-18 (15+ digits) ↗
Item consistency (Cronbach's alpha)Multivariateitems = i1, i2, i3✓ Δ 2.2×10-16 (15+ digits) ↗
Item consistency (Cronbach's alpha)Multivariate optionitems = i1, i2, i3; reverse items = i3; scale high = 5; scale low = 1✓ Δ 1.3×10-15 (15+ digits) ↗
Item consistency (Cronbach's alpha)Retained correctionaggregation = mean; names = scale_score✓ exact ↗
Item consistency (Cronbach's alpha)Retained scoringaggregation = mean; names = scale_score✓ Δ 4.1×10-17 (15+ digits) ↗
Linear regressionSurvey linear regression with an interaction and 90% intervalsconfidence level = 0.9; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; inference = design_df; interactions = grp:x✓ Δ 6.8×10-14 (14 digits) ↗
Linear regressionSurvey linear regression with design-df t and F inferenceconfidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; inference = design_df✓ Δ 5.5×10-13 (13 digits) ↗
Linear regressionSurvey linear regression with normal inferenceconfidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; inference = normal✓ Δ 1.7×10-12 (12 digits) ↗
Logistic regressionSurvey logistic regression with a first-stage FPC and 90% intervalsconfidence level = 0.9; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu, fpc N, fpc kind count; event value = yes; inference = design_df✓ Δ 4.6×10-11 (12 digits) ↗
Logistic regressionSurvey logistic regression with design-df inferenceconfidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; event value = yes; inference = design_df✓ Δ 1.7×10-10 (11 digits) ↗
Logistic regressionSurvey logistic regression with normal inferenceconfidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; event value = yes; inference = normal✓ Δ 5.9×10-11 (11 digits) ↗
Means, totals and proportionsSurvey domain estimates with 90% intervalsbounded interval = none; confidence level = 0.9; design = weight w, variance linearization, stage1 strata h, units psu; design effect = none; domain values = A✓ Δ 6.2×10-14 (14 digits) ↗
Means, totals and proportionsSurvey mean and proportion from supplied replicate weightsbounded interval = none; confidence level = 0.95; design = weight w, variance replicate, replicates columns rw1, rw2, rw3, rw4, method other, weight type full, scale 0.75, rsca...; design effect = none✓ Δ 8.5×10-15 (13 digits) ↗
Means, totals and proportionsSurvey mean and total with a first-stage FPCbounded interval = none; confidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu, fpc N, fpc kind count; design effect = none✓ Δ 6.8×10-13 (15+ digits) ↗
Means, totals and proportionsSurvey mean, total, proportions and counts (Taylor)bounded interval = none; confidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu; design effect = none✓ Δ 4.5×10-13 (13 digits) ↗
Means, totals and proportionsSurvey proportion with Korn-Graubard limits and SRS design effectbounded interval = korn_graubard; confidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu; design effect = srs_wr✓ Δ 4.4×10-13 (14 digits) ↗
Means, totals and proportionsSurvey proportion with a delta-logit intervalbounded interval = delta_logit; confidence level = 0.95; design = weight w, variance linearization, stage1 strata h, units psu; design effect = none✓ Δ 5.6×10-16 (14 digits) ↗
Multinomial regressionno comparison case yet
Ordinal regressionSurvey proportional-odds regression (Taylor)confidence level = 0.95; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; level order = low, mid, high✓ Δ 1.5×10-13 (14 digits) ↗
Ordinal regressionSurvey proportional-odds regression in a domain with 90% intervalsconfidence level = 0.9; covariates = x; design = weight w, variance linearization, stage1 strata h, units psu; domain values = A; level order = low, mid, high✓ Δ 9.7×10-13 (11 digits) ↗
Rater agreementRater agreement: Fleiss' kappa for three raters with missing ratingsalpha metric = nominal; categories = 1, 2, 3, 4; confidence level = 0.95; kappa bootstrap = no; raters = r1, r2, r3; replicates = 500; seed = 20260928✓ Δ 3.6×10-17 (14 digits) ↗
Rater agreementRater agreement: linear-weighted kappa and ordinal alphaalpha metric = ordinal; categories = 1, 2, 3, 4; confidence level = 0.95; kappa bootstrap = no; raters = r1, r2; replicates = 500; seed = 20260928✓ Δ 5.6×10-16 (11 digits) ↗
Rater agreementRater agreement: quadratic-weighted kappa at 90% and interval alphaalpha metric = interval; categories = 1, 2, 3, 4; confidence level = 0.9; kappa bootstrap = no; raters = r1, r2; replicates = 500; seed = 20260928✓ Δ 1.1×10-14 (12 digits) ↗
Rater agreementRater agreement: unweighted Cohen's kappa and nominal alphaalpha metric = nominal; confidence level = 0.95; kappa bootstrap = no; raters = r1, r2; replicates = 500; seed = 20260928✓ Δ 1.1×10-16 (13 digits) ↗
RatiosSurvey ratio from supplied replicate weightsconfidence level = 0.95; denominator = x; design = weight w, variance replicate, replicates columns rw1, rw2, rw3, rw4, method other, weight type full, scale 0.75, rsca...; numerators = y✓ Δ 4.5×10-15 (13 digits) ↗
RatiosSurvey ratio of weighted totals (Taylor)confidence level = 0.95; denominator = x; design = weight w, variance linearization, stage1 strata h, units psu; numerators = y✓ Δ 1.7×10-15 (14 digits) ↗
RatiosSurvey ratios of two numerators in a domainconfidence level = 0.9; denominator = x; design = weight w, variance linearization, stage1 strata h, units psu; domain values = A; numerators = y, z✓ Δ 4.9×10-16 (15+ digits) ↗
Scale reliabilityScale reliability of four items without reversal or ICCconfidence level = 0.95; icc model = none; icc type = consistency; items = i1, i2, i3, i5; split = halves✓ Δ 5.0×10-14 (14 digits) ↗
Scale reliabilityScale reliability with a declared split and one-way ICC at 90%confidence level = 0.9; first half = i2, i5; icc model = one_way; icc type = consistency; items = i1, i2, i3, i4, i5; reverse items = i4; scale high = 5; scale low = 1; split = custom✓ Δ 3.7×10-14 (13 digits) ↗
Scale reliabilityScale reliability with a reverse-coded item (halves split, ICC(3,1))confidence level = 0.95; icc model = two_way_mixed; icc type = consistency; items = i1, i2, i3, i4, i5; reverse items = i4; scale high = 5; scale low = 1; split = halves✓ Δ 3.7×10-14 (13 digits) ↗
Scale reliabilityScale reliability with an odd/even split and ICC(2,1) absolute agreementconfidence level = 0.95; icc model = two_way_random; icc type = agreement; items = i1, i2, i3, i4, i5; reverse items = i4; scale high = 5; scale low = 1; split = odd_even✓ Δ 3.7×10-14 (13 digits) ↗
Survey means and proportionsSurvey meanmode = mean✓ Δ 4.1×10-16 (15+ digits) ↗
Survey means and proportionsSurvey proportionmode = proportion✓ Δ 3.0×10-17 (15+ digits) ↗

Time Series

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
ACF & PACFBiased sample ACF and lag-one PACFmaximum lag = 5; task = correlations✓ Δ 3.4×10-16 (15+ digits) ↗
ACF & PACFNIST Lew: acfmaximum lag = 1; task = correlations✓ Δ 4.4×10-16 (14 digits) ↗
ACF & PACFNIST Lottery: acfmaximum lag = 1; task = correlations✓ Δ 1.2×10-16 (15+ digits) ↗
ACF & PACFNIST Mavro: acfmaximum lag = 1; task = correlations✓ Δ 1.7×10-14 (13 digits) ↗
ACF & PACFNIST Michelso: acfmaximum lag = 1; task = correlations✓ Δ 2.0×10-14 (13 digits) ↗
ACF & PACFNIST NumAcc2: acfmaximum lag = 1; task = correlations✓ Δ 1.1×10-16 (15+ digits) ↗
ACF & PACFNIST NumAcc3: acfmaximum lag = 1; task = correlations✓ Δ 8.9×10-19 (15+ digits) ↗
ACF & PACFNIST NumAcc4: acfmaximum lag = 1; task = correlations✓ Δ 2.2×10-16 (15+ digits) ↗
ACF & PACFNIST PiDigits: acfmaximum lag = 1; task = correlations✓ Δ 1.6×10-18 (15+ digits) ↗
Compare and validate forecastsStudyforecast = response column rw, model naive, horizon 3; validation = initial window 40, horizon 3, final evaluation 6✓ Δ 3.1×10-15 (14 digits) ↗
Cross-correlationA 1e8 offset leaves the correlations unchangedmax lag = 4✓ Δ 9.8×10-10 (8 digits) ↗
Cross-correlationA delayed copy peaks at its delaymax lag = 8✓ Δ 2.3×10-16 (15+ digits) ↗
Cross-correlationA duplicated time key is refusedmax lag = 4✓ ↗
Cross-correlationA maximum lag above 250 is refusedmax lag = 251✓ ↗
Cross-correlationFirst differences of both seriesmax lag = 5✓ Δ 4.4×10-16 (15+ digits) ↗
Cross-correlationLags with fewer than three pairs are omittedmax lag = 30✓ Δ 5.4×10-17 (14 digits) ↗
Cross-correlationMissing and excluded rows stay as unavailable positionsmax lag = 6✓ Δ 3.3×10-16 (15+ digits) ↗
Cross-correlationPrewhitening with an OLS AR(2) fitmax lag = 5; prewhiten = yes; prewhiten order = 2✓ Δ 3.7×10-16 (13 digits) ↗
Cross-correlationTwo rows give an insufficient-data outcomemax lag = 1✓ exact ↗
Dynamic regressionDynamic regressionarima order = 1, 0, 0; deterministic term = constant; features = column x, kind level; future scenarios = column x, values 4, 5, 6✓ Δ 2.5×10-8 (7 digits) ↗
Fit ARMA or ARIMAForecastarima order = 1, 0, 0; deterministic term = constant; horizon = 3✓ Δ 2.2×10-6 (6 digits) ↗
Fit ARMA or ARIMAForecastarima order = 1, 1, 0; deterministic term = none; horizon = 3✓ Δ 1.5×10-9 (7 digits) ↗
ForecastForecasthorizon = 3✓ Δ 7.5×10-15 (15+ digits) ↗
ForecastForecasthorizon = 3; season length = 4✓ Δ 4.0×10-15 (15+ digits) ↗
GARCH modelsAdvanceddistribution = normal; garch p = 1; garch q = 1; horizon = 1; kind = garch; mean = constant✓ Δ 1.1×10-8 (7 digits) ↗
Seasonal decomposition and adjustmentA random walk is a differencing candidatetask = stationarity✓ exact ↗
Seasonal decomposition and adjustmentAdvancedbreak deterministic = both; kind = zivot_andrews✓ Δ 5.9×10-13 (13 digits) ↗
Seasonal decomposition and adjustmentAdvancedcointegration rank = 0; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 1✓ Δ 4.3×10-17 (15+ digits) ↗
Seasonal decomposition and adjustmentAdvancedcointegration rank = 1; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 2✓ Δ 4.1×10-11 (12 digits) ↗
Seasonal decomposition and adjustmentAdvancedcointegration rank = 2; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 1✓ Δ 1.8×10-15 (13 digits) ↗
Seasonal decomposition and adjustmentAdvanceddistribution = normal; garch p = 1; garch q = 1; horizon = 1; kind = garch; mean = constant✓ Δ 1.1×10-8 (7 digits) ↗
Seasonal decomposition and adjustmentBiased sample ACF and lag-one PACFmaximum lag = 5; task = correlations✓ Δ 3.4×10-16 (15+ digits) ↗
Seasonal decomposition and adjustmentDynamic regressionarima order = 1, 0, 0; deterministic term = constant; features = column x, kind level; future scenarios = column x, values 4, 5, 6✓ Δ 2.5×10-8 (7 digits) ↗
Seasonal decomposition and adjustmentFirst ordinary difference of the guided workflowdifference order = 2; maximum lag = 5; task = difference✓ Δ 1.9×10-18 (15+ digits) ↗
Seasonal decomposition and adjustmentForecastarima order = 1, 0, 0; deterministic term = constant; horizon = 3✓ Δ 2.2×10-6 (6 digits) ↗
Seasonal decomposition and adjustmentForecastarima order = 1, 1, 0; deterministic term = none; horizon = 3✓ Δ 1.5×10-9 (7 digits) ↗
Seasonal decomposition and adjustmentForecasthorizon = 3✓ Δ 7.5×10-15 (15+ digits) ↗
Seasonal decomposition and adjustmentForecasthorizon = 3; season length = 4✓ Δ 4.0×10-15 (15+ digits) ↗
Seasonal decomposition and adjustmentGuided selection of the first stationary difference orderdifference order = 2; maximum lag = 5; task = difference✓ Δ 5.3×10-15 (14 digits) ↗
Seasonal decomposition and adjustmentNIST Lew: acfmaximum lag = 1; task = correlations✓ Δ 4.4×10-16 (14 digits) ↗
Seasonal decomposition and adjustmentNIST Lottery: acfmaximum lag = 1; task = correlations✓ Δ 1.2×10-16 (15+ digits) ↗
Seasonal decomposition and adjustmentNIST Mavro: acfmaximum lag = 1; task = correlations✓ Δ 1.7×10-14 (13 digits) ↗
Seasonal decomposition and adjustmentNIST Michelso: acfmaximum lag = 1; task = correlations✓ Δ 2.0×10-14 (13 digits) ↗
Seasonal decomposition and adjustmentNIST NumAcc2: acfmaximum lag = 1; task = correlations✓ Δ 1.1×10-16 (15+ digits) ↗
Seasonal decomposition and adjustmentNIST NumAcc3: acfmaximum lag = 1; task = correlations✓ Δ 8.9×10-19 (15+ digits) ↗
Seasonal decomposition and adjustmentNIST NumAcc4: acfmaximum lag = 1; task = correlations✓ Δ 2.2×10-16 (15+ digits) ↗
Seasonal decomposition and adjustmentNIST PiDigits: acfmaximum lag = 1; task = correlations✓ Δ 1.6×10-18 (15+ digits) ↗
Seasonal decomposition and adjustmentNo eligible differencing stagedifference order = 2; maximum lag = 5; task = difference✓ Δ 3.9×10-12 (11 digits) ↗
Seasonal decomposition and adjustmentOriginal series selected: no differencing recommendeddifference order = 2; maximum lag = 5; task = stationarity✓ Δ 1.5×10-14 (14 digits) ↗
Seasonal decomposition and adjustmentStudyforecast = response column rw, model naive, horizon 3; validation = initial window 40, horizon 3, final evaluation 6✓ Δ 3.1×10-15 (14 digits) ↗
Seasonal decomposition and adjustmentWhite noise is stationarytask = stationarity✓ exact ↗
Stationarity and differencingA random walk is a differencing candidatetask = stationarity✓ exact ↗
Stationarity and differencingFirst ordinary difference of the guided workflowdifference order = 2; maximum lag = 5; task = difference✓ Δ 1.9×10-18 (15+ digits) ↗
Stationarity and differencingGuided selection of the first stationary difference orderdifference order = 2; maximum lag = 5; task = difference✓ Δ 5.3×10-15 (14 digits) ↗
Stationarity and differencingNo eligible differencing stagedifference order = 2; maximum lag = 5; task = difference✓ Δ 3.9×10-12 (11 digits) ↗
Stationarity and differencingOriginal series selected: no differencing recommendeddifference order = 2; maximum lag = 5; task = stationarity✓ Δ 1.5×10-14 (14 digits) ↗
Stationarity and differencingWhite noise is stationarytask = stationarity✓ exact ↗
Trend analysisExponentialhorizon = 3✓ Δ 8.5×10-14 (13 digits) ↗
Trend analysisExponential NLShorizon = 0✓ Δ 1.0×10-9 (7 digits) ↗
Trend analysisLinearhorizon = 1✓ Δ 1.6×10-14 (14 digits) ↗
Trend analysisLinearhorizon = 2✓ Δ 1.8×10-14 (14 digits) ↗
Trend analysisQuadratichorizon = 0✓ Δ 9.2×10-14 (14 digits) ↗
Trend analysisScurvehorizon = 1✓ Δ 6.1×10-8 (7 digits) ↗
Unit-root test with a breakAdvancedbreak deterministic = both; kind = zivot_andrews✓ Δ 5.9×10-13 (13 digits) ↗
VAR and VECMAdvancedcointegration rank = 0; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 1✓ Δ 4.3×10-17 (15+ digits) ↗
VAR and VECMAdvancedcointegration rank = 1; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 2✓ Δ 4.1×10-11 (12 digits) ↗
VAR and VECMAdvancedcointegration rank = 2; deterministic = constant; horizon = 2; kind = var_vecm; lag order = 1✓ Δ 1.8×10-15 (13 digits) ↗

Quality

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Acceptance sampling by variablesKnown sigma one-sided variables planaql = 0.02; consumer risk = 0.1; ltpd = 0.1; max sample size = 500; oc points = 11; producer risk = 0.05; sigma = 2.25; sigma status = known; spec limit = 87.5; spec side = upper✓ Δ 2.2×10-16 (14 digits) ↗
Acceptance sampling by variablesUnknown sigma one-sided variables planaql = 0.02; consumer risk = 0.1; ltpd = 0.1; max sample size = 500; oc points = 11; producer risk = 0.05; sigma status = unknown; spec limit = 12.5; spec side = lower✓ exact ↗
Apply specificationsCase-sensitive matching leaves 'length' unmatchedcase insensitive = no✓ exact ↗
Apply specificationsDefault normalization matches through whitespace and casedefault settings✓ Δ 3.6×10-16 (15+ digits) ↗
Apply specificationsKeeping whitespace leaves ' Diameter A' unmatchedignore whitespace = no✓ exact ↗
Apply specificationsMatching a transposed sheettransposed = yes✓ exact ↗
Apply specificationsNeither limit role assigned is refuseddefault settings✓ ↗
Apply specificationsTransposing a sheet laid out across columnstransposed = yes✓ exact ↗
Apply specificationsWithout transposing the sheet is unchangedtransposed = no✓ exact ↗
Manufacturing decision toolsAn AQL at or above the LTPD is refusedaql = 0.05; ltpd = 0.05✓ ↗
Manufacturing decision toolsBinomial single-sampling plan: AQL 0.01, LTPD 0.06aql = 0.01; consumer risk = 0.1; ltpd = 0.06; producer risk = 0.05✓ Δ 5.3×10-16 (14 digits) ↗
Manufacturing decision toolsBinomial single-sampling plan: AQL 0.02, LTPD 0.08aql = 0.02; consumer risk = 0.05; ltpd = 0.08; producer risk = 0.1✓ Δ 3.3×10-16 (14 digits) ↗
Manufacturing decision toolsDistribution freeconfidence = 0.9; content = 0.9; sided = one_upper✓ ↗
Manufacturing decision toolsNIST (52,3) binomial operating-characteristic exampleaccept number = 3; sample size = 52✓ Δ 1.1×10-16 (15+ digits) ↗
Manufacturing decision toolsNormalconfidence = 0.95; content = 0.95; sided = two_sided✓ Δ 5.3×10-15 (14 digits) ↗
Manufacturing decision toolsNormalconfidence = 0.95; content = 0.99; sided = one_upper✓ Δ 9.5×10-12 (12 digits) ↗
Manufacturing decision toolsNormalconfidence = 0.99; content = 0.95; sided = two_sided✓ Δ 1.6×10-15 (15+ digits) ↗
Manufacturing decision toolsRight-censored Weibull life modelfailure value = failed✓ Δ 3.9×10-5 (6 digits) ↗

Quality › Capability

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Assess capabilityBinomialconfidence level = 0.9; family = binomial✓ Δ 2.9×10-20 (15+ digits) ↗
Assess capabilityBinomialfamily = binomial✓ Δ 3.9×10-13 (15+ digits) ↗
Assess capabilityNormallimits = value 5.49, 5.79; study = subgroup mode constant, subgroup size 5, within method pooled, require stability no, require msa no✓ Δ 7.1×10-15 (14 digits) ↗
Assess capabilityNormallimits = value 5.49, 5.79; study = subgroup mode constant, subgroup size 5, within method range, require stability no, require msa no✓ Δ 6.9×10-15 (14 digits) ↗
Assess capabilityNormallimits = value 5.49, 5.79; study = subgroup mode constant, subgroup size 5, within method stddev, require stability no, require msa no✓ Δ 7.1×10-15 (14 digits) ↗
Assess capabilityNormallimits = value 5.49, 5.79; study = subgroup mode individuals, within method moving_range, require stability no, require msa no✓ Δ 6.9×10-15 (14 digits) ↗
Assess capabilityNormallimits = value 8.7, 11.3; study = subgroup mode individuals, within method moving_range, require stability no, require msa no✓ Δ 2.7×10-15 (14 digits) ↗
Assess capabilityPoissonfamily = poisson✓ Δ 2.6×10-15 (15+ digits) ↗
Assess capabilitySBa = 2.373; b = 1.959; family = SB; loc = 0.295; scale = 1.203✓ Δ 2.6×10-17 (15+ digits) ↗
Assess capabilitySUa = 1.402; b = 2.333; family = SU; loc = 15.516; scale = 1.585✓ Δ 2.7×10-16 (15+ digits) ↗
Capability of several characteristicsNormalconfidence level = 0.95; lsls = 8.8, 1.75; targets = 10, ; usls = 11.2, 2.25✓ Δ 7.3×10-11 (11 digits) ↗
Capability of several characteristicsNormalconfidence level = 0.99; lsls = , 1.75; usls = 11.2, ✓ Δ 7.2×10-11 (11 digits) ↗
Capability sixpackNormalconfidence level = 0.95; lsl = 8.8; target = 10; usl = 11.2✓ Δ 4.7×10-11 (11 digits) ↗
Capability sixpackNormalconfidence level = 0.95; usl = 2.25✓ Δ 5.6×10-11 (13 digits) ↗
Capability sixpackNormalconfidence level = 0.9; lsl = 8.8; target = 10; usl = 11.2✓ Δ 4.7×10-11 (11 digits) ↗
Compare capability of populationsNormalconfidence level = 0.95; lsl = 8.8; split kind = group; target = 10; usl = 11.2✓ Δ 9.2×10-11 (9 digits) ↗
Compare capability of populationsNormalconfidence level = 0.95; lsl = 8.8; split kind = point; split value = 21; usl = 11.2✓ Δ 9.2×10-11 (9 digits) ↗
Within/between batch capabilityA zero between-batch component makes within and overall coincidelower limit = 46; upper limit = 54✓ Δ 1.1×10-16 (15+ digits) ↗
Within/between batch capabilityNo specification limit is refuseddefault settings✓ ↗
Within/between batch capabilityUpper limit only at a 90% levelconfidence level = 0.9; upper limit = 54✓ Δ 3.1×10-15 (14 digits) ↗
Within/between batch capabilityWithin and overall capability of balanced batcheslower limit = 46; upper limit = 54✓ Δ 2.1×10-14 (14 digits) ↗

Quality › Measurement systems

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Attribute agreementRepeated columns are refusedappraiser = item; part = item; rating = rating✓ ↗
Attribute agreementThree appraisers, two trials and a known standardappraiser = appraiser; part = item; rating = rating; study standard supplied; trial = trial✓ Δ 1.5×10-14 (15+ digits) ↗
Attribute agreementTwo appraisers use Cohen kappa between themappraiser = appraiser; part = item; rating = rating; trial = trial✓ Δ 1.9×10-16 (15+ digits) ↗
Attribute agreementWeighted Cohen kappa: ordinal linearappraiser = appraiser; part = item; rating = rating; scale = ordinal_linear✓ Δ 1.1×10-16 (15+ digits) ↗
Attribute agreementWeighted Cohen kappa: ordinal quadraticappraiser = appraiser; part = item; rating = rating; scale = ordinal_quadratic✓ Δ 5.5×10-16 (14 digits) ↗
Attribute gage response curve10/90 transition points at 90% confidenceconfidence level = 0.9; lower percentile = 0.1; threshold = 10; upper percentile = 0.9✓ Δ 2.1×10-9 (7 digits) ↗
Attribute gage response curveModelling the fail event reverses the curvemodeled event = fail; threshold = 10✓ Δ 2.5×10-9 (7 digits) ↗
Attribute gage response curveProbit gage response with Fieller intervalsthreshold = 10✓ Δ 1.6×10-7 (5 digits) ↗
Attribute gage response curveSeparated outcomes have no finite estimatethreshold = 2.5✓ ↗
Extended gage studiesFewer than ten measurements are refusedreference = 10; tolerance = 0.6✓ ↗
Extended gage studiesGage linearity and biastolerance = 4✓ Δ 2.2×10-15 (13 digits) ↗
Extended gage studiesGage stability sorted by an order columnreference = 5.1✓ Δ 8.9×10-16 (13 digits) ↗
Extended gage studiesNested destructive Gage R&Rdefault settings✓ Δ 2.8×10-14 (14 digits) ↗
Extended gage studiesType-1 gage Cg and Cgkreference = 10; tolerance = 0.6✓ Δ 4.5×10-14 (12 digits) ↗
Gage R repeatability screeningANOVAmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance 8, intended use processnot assessed
Gage R repeatability screeningAverage rangemax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R repeatability screeningChart worksheetmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R repeatability screeningPooled SDmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R repeatability screeningRangemax rows = 1000000; process sd = 0.0777; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R&RA missing value unbalances the design and is refuseddefault settings✓ ↗
Gage R&RAIAG MSA-4 Appendix B impact of GRR on Cp (corrected tolerance formula)measurement sd = 0.6; observed cp = 2; observed sd = 1; tolerance = 12✓ exact ↗
Gage R&RANOVAmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance 8, intended use processnot assessed
Gage R&RAverage rangemax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R&RChart worksheetmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R&RCrossed Gage R&R with a pooled interaction and a tolerancetolerance = 12✓ Δ 1.3×10-13 (13 digits) ↗
Gage R&RCrossed Gage R&R with a retained interactiondefault settings✓ Δ 8.2×10-14 (13 digits) ↗
Gage R&RInteraction alpha 0.99 and a 5.15-sigma study variationinteraction alpha = 0.99; study var multiplier = 5.15✓ Δ 1.3×10-13 (13 digits) ↗
Gage R&ROne trial per cell is refuseddefault settings✓ ↗
Gage R&RPooled SDmax rows = 1000000; process sd = None; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R&RRangemax rows = 1000000; process sd = 0.0777; study standard supplied; total variation = option study, process spread None, target pp None, tolerance None, intended use processnot assessed
Gage R&RSimultaneous full-model repeatability intervaldefault settings✓ Δ 1.2×10-15 (14 digits) ↗
Gage run chartGage run chart means and trial layout on a hand-checked crossed studydefault settings✓ Δ 3.6×10-16 (15+ digits) ↗
Multifactor gage R&RA part-by-site interaction is confounded when sites nest in partsinteractions = part_site; structure = site_nested_in_part✓ ↗
Multifactor gage R&RClassic crossed structure reproduces the crossed enginedefault settings✓ Δ 5.3×10-15 (14 digits) ↗
Multifactor gage R&RFully crossed part, operator and sitestructure = fully_crossed✓ Δ 2.2×10-7 (6 digits) ↗
Multifactor gage R&RSite nested in part counts toward part-to-partstructure = site_nested_in_part✓ Δ 1.1×10-7 (6 digits) ↗

Quality › Reliability engineering

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Accelerated life analysisAccelerated life analysishigh stress = 104000; mode = accelerated; stress type = linear; stress unit = native; use stress = 102500✓ Δ 8.4×10-8 (13 digits) ↗
Accelerated life analysisAccelerated life analysishigh stress = 60; mode = accelerated; stress type = linear; stress unit = native; use stress = 40✓ exact ↗
Accelerated life analysisParametric survivalhigh stress = 80; mode = accelerated; stress type = linear; stress unit = native; use stress = 50✓ Δ 2.8×10-12 (12 digits) ↗
Accelerated life analysisSurvival corrected optionconfidence level = 0.9; high stress = 80; mode = accelerated; stress type = inverse_power; stress unit = native; use stress = 50✓ Δ 1.0×10-11 (12 digits) ↗
Accelerated life analysisSurvival optionconfidence level = 0.9; high stress = 80; mode = accelerated; stress type = arrhenius; stress unit = celsius; use stress = 25✓ Δ 2.3×10-10 (10 digits) ↗
Accelerated life analysisSurvival optionconfidence level = 0.9; high stress = 80; mode = accelerated; stress type = inverse_power; stress unit = native; use stress = 50✓ Δ 7.2×10-12 (12 digits) ↗
Accelerated life analysisSurvival populationhigh stress = 80; mode = accelerated; stress type = linear; stress unit = native; use stress = 50✓ exact ↗
Censored lifetime distributionsCensored lifetime distributionsdefault settings✓ ↗
Censored lifetime distributionsCensored lifetime distributionsmode = families✓ Δ 1.1×10-11 (15+ digits) ↗
Censored lifetime distributionsCensored lifetime distributionsmode = unsupported✓ ↗
Censored lifetime distributionsParametric survivalmode = families✓ Δ 2.5×10-16 (15+ digits) ↗
Censored lifetime distributionsSurvival populationmode = families✓ Δ 1.3×10-16 (15+ digits) ↗
Degradation-based lifetimeDegradation-based lifetime: inverse-kelvin temperature acceleration to a use temperaturebaseline = 0; confidence level = 0.95; direction = increasing; dropout ignorable declared = yes; lifetime probabilities = 0.1, 0.5, 0.9; mode = repeated; reliability times = 10, 15, 20; threshold = 25; time unit = hours; use temperature = 348not assessed
Degradation-based lifetimeDegradation-based lifetime: repeated measurements, lognormal rates, no temperature axisbaseline = 0; confidence level = 0.95; direction = increasing; dropout ignorable declared = yes; lifetime probabilities = 0.1, 0.5, 0.9; mode = repeated; reliability times = 10, 15, 20; threshold = 25; time unit = hours; use temperature = Nonenot assessed
Distribution identification plotDistribution identification of complete data across the seven familiesconfidence level = 0.95; end = time; families = weibull, lognormal, exponential, loglogistic, logistic, normal, sev; frequency = freq; layout = arbitrary; start = time; threshold = 0✓ Δ 7.7×10-6 (6 digits) ↗
Distribution identification plotDistribution identification of right-censored data with the restricted-domain ADconfidence level = 0.95; families = weibull, lognormal, exponential, loglogistic, logistic, normal, sev; frequency = freq; layout = right_censored; start = time; status = status; threshold = 0✓ Δ 1.8×10-5 (6 digits) ↗
Mixed-censoring life analysisMixed-censoring life analysismode = interval✓ Δ 3.1×10-6 (10 digits) ↗
Mixed-censoring life analysisParametric survivalmode = interval✓ Δ 6.5×10-8 (7 digits) ↗
Mixed-censoring life analysisSurvival populationmode = interval✓ exact ↗
Nonparametric life estimationKaplan-Meier life estimation with a tied failure and censoring at 90%confidence level = 0.9; frequency = freq; layout = right_censored; start = time; status = status✓ Δ 1.4×10-15 (15+ digits) ↗
Nonparametric life estimationKaplan-Meier life estimation with frequencies at 95%confidence level = 0.95; frequency = freq; layout = right_censored; start = time; status = status✓ Δ 1.4×10-15 (14 digits) ↗
Nonparametric life estimationTurnbull nonparametric life estimation from interval dataconfidence level = 0.95; end = end; frequency = freq; layout = arbitrary; start = start✓ Δ 1.5×10-15 (14 digits) ↗
Parametric life distributionExponentialconfidence level = 0.95; family = exponential; frequency = freq; layout = right_censored; percents = 1, 5, 50; start = start; status = status; threshold = 0✓ Δ 5.0×10-6 (6 digits) ↗
Parametric life distributionLognormalconfidence level = 0.9; end = end; family = lognormal; frequency = freq; layout = arbitrary; percents = 5, 50; start = start; threshold = 5✓ Δ 1.4×10-7 (7 digits) ↗
Parametric life distributionSEVconfidence level = 0.95; end = end; family = sev; frequency = freq; layout = arbitrary; percents = 1, 10, 50; start = start; threshold = 0✓ Δ 1.5×10-5 (6 digits) ↗
Parametric life distributionWeibullconfidence level = 0.95; end = end; family = weibull; frequency = freq; layout = arbitrary; percents = 1, 5, 10, 50; start = start; threshold = 0✓ Δ 2.2×10-6 (6 digits) ↗
Prepare warranty dataDerived warranty censoring table with interval records and right-censored survivorslast period = incomplete; period unit = month; period width = 1; return axis = calendar; returns = r1, r2, r3, r4, r5, r6, r7; shipment label = label; shipped = shipped✓ exact ↗
Prepare warranty dataWarranty triangle conversion on the age axis with the final period completelast period = complete; period unit = month; period width = 1; return axis = age; returns = r1, r2, r3, r4, r5; shipment label = label; shipped = shipped✓ exact ↗
Prepare warranty dataWarranty triangle conversion on the calendar axis with the final period incompletelast period = incomplete; period unit = month; period width = 1; return axis = calendar; returns = r1, r2, r3, r4, r5, r6, r7; shipment label = label; shipped = shipped✓ exact ↗
Regression with life dataLognormalconfidence level = 0.95; end = end; family = lognormal; frequency = freq; layout = arbitrary; percents = 10, 50; prediction rows = temp3=low, temp3=high; start = start; temperature unit = celsius; threshold = 0; transform = none✓ Δ 1.6×10-6 (6 digits) ↗
Regression with life dataSEVconfidence level = 0.9; end = end; family = sev; frequency = freq; layout = arbitrary; percents = 10, 90; prediction rows = temp_c=45, temp_c=75; start = start; temperature unit = celsius; threshold = 0; transform = arrhenius✓ Δ 7.3×10-3 (5 digits) ↗
Regression with life dataWeibullconfidence level = 0.95; end = end; family = weibull; frequency = freq; layout = arbitrary; percents = 10, 50; prediction rows = stress=80, stress=150; start = start; temperature unit = celsius; threshold = 0; transform = none✓ Δ 1.4×10-6 (7 digits) ↗
Repairable systemsRepairable-system MCF with a constant-rate HPP overlayconfidence level = 0.95; evaluation ages = 3, 6, 12; events tab = events; stored fits supplied; time unit = months; windows tab = windows✓ Δ 2.3×10-15 (15+ digits) ↗
Repairable systemsRepairable-system MCF with a power-law NHPP fit and forecastconfidence level = 0.9; evaluation ages = 4, 8; events tab = events; forecast horizon = 2; stored fits supplied; time unit = months; windows tab = windows✓ Δ 5.8×10-8 (7 digits) ↗
Stability studyStability studydefault settings✓ ↗
Stability studyStability studylsl = 1000✓ exact ↗
Stability studyStability studylsl = 100; usl = 95✓ ↗
Stability studyStability studylsl = 95✓ Δ 1.3×10-13 (11 digits) ↗
Stability studySurvivallsl = 95; model type = pooled✓ Δ 3.3×10-14 (14 digits) ↗
Stability studySurvival optionconfidence level = 0.9; lsl = 95; model type = separate_intercepts_pooled_slope✓ Δ 2.3×10-14 (13 digits) ↗
Stability studySurvival optionlsl = 95; model type = separate_intercepts_separate_slopes✓ Δ 3.4×10-14 (13 digits) ↗
Stability studySurvival populationlsl = 95✓ Δ 6.7×10-16 (14 digits) ↗
Warranty predictionLognormalconfidence level = 0.9; end = end; family = lognormal; frequency = freq; future shipments = 400@2.5; horizon = 8; period unit = month; period width = 1; source = censoring_table; start = start✓ Δ 2.5×10-7 (7 digits) ↗
Warranty predictionWeibullconfidence level = 0.95; family = weibull; future shipments = 500@1.5, 300@6; horizon = 12; last period = incomplete; period unit = month; period width = 1; return axis = calendar; returns = r1, r2, r3, r4, r5, r6, r7; shipped = shipped; source = triangle; unit cost = 3.2; warranty limit = 24✓ Δ 1.3×10-5 (7 digits) ↗
Weibull accelerated failure timeParametric survivalprediction stresses = 50✓ Δ 1.4×10-9 (9 digits) ↗
Weibull accelerated failure timeSurvival populationprediction stresses = 50✓ exact ↗
Weibull accelerated failure timeWeibull accelerated failure timedefault settings✓ exact ↗✓ ↗
Weibull accelerated failure timeWeibull accelerated failure timemax rows = 0✓ ↗
Weibull accelerated failure timeWeibull accelerated failure timeprediction stresses = 102500✓ Δ 2.0×10-7 (10 digits) ↗

Quality › Statistical process control

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Between/within I-MR-R/S chartBetween/within I-MR-R chart by subgroup columndispersion = range✓ Δ 1.1×10-14 (14 digits) ↗
Between/within I-MR-R/S chartBetween/within I-MR-S chartdispersion = stddev✓ Δ 1.1×10-14 (14 digits) ↗
Between/within I-MR-R/S chartBetween/within chart with consecutive subgroups of five and two stagesdispersion = range✓ Δ 1.2×10-14 (14 digits) ↗
Between/within I-MR-R/S chartBetween/within chart with unequal lot sizes and an excluded rowdispersion = range✓ Δ 1.5×10-8 (7 digits) ↗
Control chartsA non-positive sigma is refusedsigma = 0✓ ↗
Control chartsA subgroup size of 26 is refuseddefault settings✓ ↗
Control chartsAn unknown run rule is refusedrun rules = rule_9✓ ↗
Control chartsApproved limits for the wrong chart are refusedapproved = chart xbar, center 21, lower 18, upper 24✓ ↗
Control chartsBenneyan geometric support translated to HypoNum g chartbaseline = column phase, op eq, value reference; chart = g✓ Δ 1.6×10-16 (15+ digits) ↗
Control chartsCUSUMcusum h = 1; cusum k = 0.5; sigma = 1; target = 0✓ exact ↗
Control chartsCUSUM with a supplied target and sigmacusum h = 4; cusum k = 0.25; sigma = 1.1; target = 10✓ Δ 9.2×10-16 (15+ digits) ↗
Control chartsCUSUM with target and sigma from the datadefault settings✓ Δ 2.6×10-15 (15+ digits) ↗
Control chartsDefectives above the sample size are refusedchart = p✓ ↗
Control chartsEWMAdefault settings✓ Δ 4.6×10-16 (15+ digits) ↗
Control chartsEWMAewma l = 2.7; ewma lambda = 0.1; sigma = 1; target = 10✓ Δ 4.7×10-16 (15+ digits) ↗
Control chartsEWMAewma l = 2.7; ewma lambda = 0.1; target = 10✓ Δ 2.1×10-16 (15+ digits) ↗
Control chartsEWMAewma l = 3; ewma lambda = 0.25; sigma = 1; target = 0✓ Δ 3.7×10-17 (15+ digits) ↗
Control chartsEWMAewma lambda = 1.5✓ ↗
Control chartsFixed approved limits replace the calculated onesapproved = chart individuals, center 21.5, lower 18.4, upper 24.6, chart moving_range, center 1, lower 0, upper 3.3✓ Δ 4.2×10-16 (15+ digits) ↗
Control chartsGrouping column with a median size above 10 uses X-bar/Sdefault settings✓ Δ 4.3×10-16 (15+ digits) ↗
Control chartsGrouping column with varying sizes (X-bar/R, point-specific limits)default settings✓ Δ 3.3×10-14 (14 digits) ↗
Control chartsIndividuals and moving range with the default rulesdefault settings✓ Δ 4.6×10-16 (15+ digits) ↗
Control chartsIndividuals chart at a 1e7 offsetdefault settings✓ Δ 1.8×10-9 (10 digits) ↗
Control chartsIndividuals chart with all eight rulesrun rules = rule_2, rule_3, rule_4, rule_5, rule_6, rule_7, rule_8✓ exact ↗
Control chartsLaney p' chartchart = laney_p✓ Δ 1.8×10-19 (15+ digits) ↗
Control chartsLaney u' chartchart = laney_u✓ Δ 1.8×10-15 (15+ digits) ↗
Control chartsMissing and excluded rows are omitted before the moving rangesdefault settings✓ Δ 7.4×10-15 (15+ digits) ↗
Control chartsMissing counts are omitted and excluded lots removedchart = p✓ Δ 2.9×10-19 (15+ digits) ↗
Control chartsNIST R-chart D4(5): printed table versus full precisiondefault settings✓ Δ 1.8×10-15 (15+ digits) ↗
Control chartsNIST wafer c chart with strict on-limit decisionchart = c✓ exact ↗
Control chartsNp with a varying size column is refusedchart = np✓ ↗
Control chartsPhase I baseline freezes the limits from the first 40 pointsbaseline = column phase, op le, value 1✓ Δ 4.9×10-16 (15+ digits) ↗
Control chartsSpec limits are overlays onlylower spec = 40; upper spec = 60✓ exact ↗
Control chartsTwo process stages recalculated independentlystages = name Before, start label P01, end label P30, name After, start label P31✓ exact ↗
Control chartsX-bar and R with constant subgroups of 5default settings✓ Δ 2.9×10-14 (14 digits) ↗
Control chartsX-bar and R with the smallest subgroups of 2default settings✓ Δ 4.5×10-17 (15+ digits) ↗
Control chartsX-bar and S with constant subgroups of 12default settings✓ Δ 2.5×10-15 (13 digits) ↗
Control chartsc chartchart = c✓ Δ 1.3×10-16 (15+ digits) ↗
Control chartsg chart plots the counts between eventschart = g✓ exact ↗
Control chartsnp chart with a constant size of 100chart = np; sample size = 100✓ Δ 1.3×10-17 (15+ digits) ↗
Control chartsp chart with a Phase I baseline of 20 lotsbaseline = column phase, op le, value 1; chart = p✓ Δ 3.8×10-19 (15+ digits) ↗
Control chartsp chart with varying inspected countschart = p✓ Δ 1.8×10-19 (15+ digits) ↗
Control chartsu chart with an area columnchart = u✓ Δ 1.8×10-15 (15+ digits) ↗
Individual Hotelling T² chartExcluded and incomplete rows leave the baseline and the monitoringalpha = 0.0027; reference value = ref✓ Δ 1.3×10-14 (14 digits) ↗
Individual Hotelling T² chartIndividual Phase II T-squared at alpha 0.0027alpha = 0.0027; reference value = ref✓ Δ 1.0×10-13 (14 digits) ↗
Individual Hotelling T² chartIndividual Phase II T-squared at alpha 0.01alpha = 0.01; reference value = ref✓ Δ 1.0×10-13 (14 digits) ↗
Individual Hotelling T² chartThree reference rows for three variables are too fewreference value = ref✓ ↗
Moving-average control chartA constant baseline is rejecteddefault settings✓ ↗
Moving-average control chartA historical mean without a sigma is refusedhistorical mean = 12✓ ↗
Moving-average control chartA stage column resets the windowhistorical mean = 12; historical sigma = 0.8; window width = 3✓ Δ 4.0×10-16 (15+ digits) ↗
Moving-average control chartFull-window startup with historical parametershistorical mean = 12; historical sigma = 0.8; partial window policy = full_window; window width = 4✓ Δ 2.9×10-16 (15+ digits) ↗
Moving-average control chartMissing and excluded rows are omitted before the windowsdefault settings✓ Δ 3.5×10-15 (15+ digits) ↗
Moving-average control chartMontgomery Example 9.3 all 30 moving averageshistorical mean = 10; historical sigma = 1; window width = 5✓ Δ 2.1×10-15 (15+ digits) ↗
Moving-average control chartWidth 3 with a Phase I marker and the moving-range sigmadefault settings✓ Δ 1.5×10-15 (15+ digits) ↗
Moving-average control chartWidth 5 with the sample standard deviationsigma estimator = stddev; window width = 5✓ Δ 1.8×10-15 (15+ digits) ↗
Moving-average control chartWidth one is the individuals specificationwindow width = 1✓ Δ 3.0×10-16 (15+ digits) ↗
Multivariate EWMA chartA smoothing weight above 1 is refuseddecision limit = 11; reference value = ref; smoothing = 1.2✓ ↗
Multivariate EWMA chartA stage column restarts the recursiondecision limit = 12; reference value = ref; smoothing = 0.3✓ Δ 8.5×10-14 (14 digits) ↗
Multivariate EWMA chartA user-calibrated limit needs a source notedecision limit = 11; limit provenance = user_calibrated; reference value = ref✓ ↗
Multivariate EWMA chartLambda 1 reproduces the individual Hotelling statisticdecision limit = 11; reference value = ref; smoothing = 1✓ Δ 1.0×10-13 (14 digits) ↗
Multivariate EWMA chartLowry Table 4 MEWMA from printed inputsdecision limit = 8.79; reference value = reference; smoothing = 0.1✓ Δ 1.0×10-14 (15+ digits) ↗
Multivariate EWMA chartMEWMA with lambda 0.2 and a trial limit of 11decision limit = 11; reference value = ref✓ Δ 6.5×10-13 (13 digits) ↗
Short-run Z-MR chartA product without parameters is an input errorproducts = product A, center 10, sigma 0.5, product B, center 25, sigma 1.2✓ ↗
Short-run Z-MR chartA stage column breaks adjacencyproducts = product A, center 10, sigma 0.5, product B, center 25, sigma 1.2, product C, center 3, sigma 0.1✓ Δ 3.8×10-17 (15+ digits) ↗
Short-run Z-MR chartA zero sigma is refusedproducts = product A, center 10, sigma 0, product B, center 25, sigma 1.2, product C, center 3, sigma 0.1✓ ↗
Short-run Z-MR chartMinitab Z-MR printed factors versus full precisionproducts = product A, center 0, sigma 1✓ Δ 3.8×10-17 (15+ digits) ↗
Short-run Z-MR chartMissing and excluded rows break adjacency but stay in the tableproducts = product A, center 10, sigma 0.5, product B, center 25, sigma 1.2, product C, center 3, sigma 0.1✓ Δ 2.5×10-18 (15+ digits) ↗
Short-run Z-MR chartReset the moving range at each product changeproducts = product A, center 10, sigma 0.5, product B, center 25, sigma 1.2, product C, center 3, sigma 0.1; reset on product change = yes✓ Δ 2.9×10-17 (15+ digits) ↗
Short-run Z-MR chartStandardized Z-MR across three productsproducts = product A, center 10, sigma 0.5, product B, center 25, sigma 1.2, product C, center 3, sigma 0.1✓ Δ 3.6×10-15 (14 digits) ↗
Subgroup T² and generalized varianceA subgroup size of 1 is refusedreference value = ref✓ ↗
Subgroup T² and generalized varianceMontgomery Example 11.1 subgroup Hotelling full-precision limitalpha = 0.001; reference value = reference✓ Δ 9.9×10-12 (12 digits) ↗
Subgroup T² and generalized varianceN <= p keeps the location chart and withholds the spread chartreference value = ref✓ Δ 6.9×10-15 (15+ digits) ↗
Subgroup T² and generalized varianceSubgroup T-squared and generalized variancereference value = ref✓ Δ 1.5×10-14 (14 digits) ↗
Subgroup T² and generalized varianceSubgroups larger than n stop the runreference value = ref✓ ↗
Zone chartRaw-individual zone chart with a missing observationstatistic = raw_individuals; units = g✓ Δ 2.2×10-16 (15+ digits) ↗
Zone chartZone chart as an individuals chart of subgroup meansstatistic = individuals_of_means✓ Δ 1.2×10-14 (14 digits) ↗
Zone chartZone chart of subgroup means with R-bar within sigmastatistic = subgroup_mean; within sigma = rbar✓ Δ 1.2×10-14 (14 digits) ↗
Zone chartZone chart of subgroup means with S-bar within sigmastatistic = subgroup_mean; within sigma = sbar✓ Δ 7.6×10-15 (14 digits) ↗
Zone chartZone chart with custom weights, reset after signal and two stagesreset after signal = yes; signal score = 6; statistic = subgroup_mean; within sigma = rbar✓ Δ 8.9×10-15 (12 digits) ↗

Design

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Create MSA study worksheetAttribute agreement worksheet with standardslayout = attribute_agreement; operator labels = Ann, Bob; part labels = P1, P2, P3; standard values = G, NG, G; trials = 1✓ exact ↗
Create MSA study worksheetCrossed worksheet: 3 parts x 2 operators x 2 trialslayout = crossed; operator labels = Ann, Bob; part labels = P1, P2, P3; seed = 11; trials = 2✓ exact ↗
Create MSA study worksheetNested worksheet: two parts per operatorlayout = nested; operator labels = Ann, Bob, Cy; part count = 2; seed = 5; trials = 3✓ exact ↗
Create MSA study worksheetPart labels are refused for nested designslayout = nested; operator count = 2; part labels = P1, P2; trials = 2✓ ↗
Create MSA study worksheetThe recorded operation recreates the worksheetlayout = crossed; operator labels = Ann, Bob; part labels = P1, P2, P3; seed = 11; trials = 2✓ exact ↗
Create MSA study worksheetThe same seed reproduces the sheetlayout = crossed; operator labels = Ann, Bob; part labels = P1, P2, P3; seed = 11; trials = 2✓ exact ↗

Design › Design of experiments (DOE)

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Analyze Taguchi designTAGUCHI-ANALYSIS 1anova summary = mean; predict levels = 2, 1; predict terms = A, B; sn = larger✓ Δ 2.8×10-14 (14 digits) ↗
Analyze Taguchi designTAGUCHI-ANALYSIS 2anova summary = mean; predict levels = 2, 1; predict terms = A, B; sn = smaller✓ Δ 2.8×10-14 (14 digits) ↗
Analyze Taguchi designTAGUCHI-ANALYSIS 3anova summary = mean; predict levels = 2, 1; predict terms = A, B; sn = nominal_variance✓ Δ 2.8×10-14 (14 digits) ↗
Analyze Taguchi designTAGUCHI-ANALYSIS 4anova summary = mean; predict levels = 2, 1; predict terms = A, B; sn = nominal_mean_sd✓ Δ 2.8×10-14 (14 digits) ↗
Analyze Taguchi designTAGUCHI-ANALYSIS 5anova summary = mean; dynamic reference = intercept; predict levels = 2, 1; predict terms = A, B; signal values = 1, 2, 3; sn = dynamic✓ Δ 7.8×10-14 (13 digits) ↗
Analyze Taguchi designTAGUCHI-ANALYSIS 6anova summary = mean; dynamic reference = origin; predict levels = 2, 1; predict terms = A, B; signal values = 1, 2, 3; sn = dynamic✓ Δ 2.8×10-14 (14 digits) ↗
Design of experiments (DOE)A factor with low above high is refuseddefault settings✓ ↗
Design of experiments (DOE)Binary response with a logit linkdesign = factors name A, low 10, high 20, name B, low 1, high 3, name C, low 100, high 200, response name event, model main_ef...✓ Δ 9.0×10-12 (9 digits) ↗
Design of experiments (DOE)Binary response with a probit linkdesign = factors name A, low 10, high 20, name B, low 1, high 3, name C, low 100, high 200, response name event, model main_ef...✓ Δ 3.7×10-9 (7 digits) ↗
Design of experiments (DOE)Centercenter count = 0; design = family full_factorial, factors name X, low 0, high 10, name Y, low 50, high 70, response name y; kind = center✓ exact ↗
Design of experiments (DOE)Centercenter count = 2; design = family full_factorial, factors name X, low 0, high 10, name Y, low 50, high 70, response name y; kind = center✓ Δ 1.1×10-17 (15+ digits) ↗
Design of experiments (DOE)Central compositeaxial type = face_centered; center points = 3; family = central_composite✓ exact ↗
Design of experiments (DOE)Existing-table factorial with two-factor interactionsdesign = factors name A, low 10, high 20, name B, low 1, high 3, name C, low 100, high 200, response name y, model interactions✓ Δ 1.1×10-14 (13 digits) ↗
Design of experiments (DOE)Full 2^3 factorial with two replicates and three centre pointscenter points = 3; replicates = 2; seed = 7✓ exact ↗
Design of experiments (DOE)Full factorialcenter points = 2; family = full_factorial✓ exact ↗
Design of experiments (DOE)Legacy factorial analysis with interactionsdefault settings✓ Δ 1.1×10-14 (13 digits) ↗
Design of experiments (DOE)Log-variance model slopes with equal replicationdesign = factors name A, low 10, high 20, name B, low 1, high 3, response name y, model main_effects✓ Δ 2.2×10-16 (15+ digits) ↗
Design of experiments (DOE)Main-effects model at 90% confidenceconfidence level = 0.9; design = factors name A, low 10, high 20, name B, low 1, high 3, name C, low 100, high 200, response name y, model main_effects✓ Δ 4.0×10-15 (14 digits) ↗
Design of experiments (DOE)The smallest design: two factors, one replicaterandomize = no✓ exact ↗
Design of experiments (DOE)Two maximized responses that both peak at the high cornerdesign = family full_factorial, factors name A, low 10, high 20, name B, low 1, high 3, name C, low 100, high 200, replicates 2; goals = response y2, goal maximize, weight 1, response y3, goal maximize, weight 1✓ exact ↗
Robust parameter designROBUST-ANALYSIS 1bootstrap replicates = 99; bootstrap seed = 11; design = factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5, high probabili...; objective = target; target = 10✓ Δ 9.3×10-14 (13 digits) ↗
Robust parameter designROBUST-ANALYSIS 2bootstrap replicates = 99; bootstrap seed = 11; design = factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.75, high probabil...; objective = minimize; target = 10✓ Δ 9.3×10-14 (13 digits) ↗
Robust parameter designROBUST-ANALYSIS 3bootstrap replicates = 99; bootstrap seed = 11; design = factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5, high probabili...; objective = maximize; target = 10✓ Δ 9.3×10-14 (13 digits) ↗
Robust parameter designROBUST-CONFIRM 1analysis = design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5, high pr...; predicted mean = 8; predicted objective = 9; runs per condition = 2; seed = 23; selected = factor C, value -1✓ Δ 1.1×10-17 (15+ digits) ↗
Robust parameter designROBUST-CONFIRM 2analysis = design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.75, high p...; predicted mean = 8; predicted objective = 9; runs per condition = 2; seed = 23; selected = factor C, value -1✓ Δ 1.3×10-17 (15+ digits) ↗
Robust parameter designROBUST-CONFIRM 3analysis = design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5, high pr...; predicted mean = 8; predicted objective = 9; runs per condition = 2; seed = 23; selected = factor C, value -1✓ Δ 9.4×10-18 (15+ digits) ↗
Robust parameter designROBUST-GENERATE 1replicates = 3; seed = 19✓ exact ↗
Robust parameter designROBUST-GENERATE 2replicates = 2; seed = 19✓ exact ↗
Robust parameter designROBUST-GENERATE 3replicates = 3; seed = 19; signal = name S, levels 1, 2, 3✓ exact ↗
Robust parameter designROBUST-PLAN 1confirmation = analysis design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5...✓ exact ↗
Robust parameter designROBUST-PLAN 2confirmation = analysis design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5...✓ exact ↗
Robust parameter designROBUST-PLAN 3confirmation = analysis design factors name C, role control, low -1, high 1, name Z, role noise, low -1, high 1, low probability 0.5...✓ exact ↗
Taguchi designTAGUCHI-DESIGN 1array = L4; randomize = no; repeats = 2✓ exact ↗
Taguchi designTAGUCHI-DESIGN 2array = L9; randomize = no; repeats = 2✓ exact ↗
Taguchi designTAGUCHI-DESIGN 3array = L4; noise = outer; outer array = L4; randomize = no; repeats = 2; signal levels = 1, 2, 3✓ exact ↗

Design › Power & sample size

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Power and sample sizeNormal variancefamily = normal_variance; sample size = 20; solve for = power; variance ratio = 2✓ Δ 4.4×10-16 (15+ digits) ↗
Power and sample sizeOne proportionanalysis = one_proportion; effect size = 0.5; sample size = 30; solve for = power✓ Δ 1.1×10-16 (15+ digits) ↗
Power and sample sizeOne proportionanalysis = one_proportion; effect size = 0.5; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizeOne sample talpha = 0.05; analysis = one_sample_t; effect size = 0.5; sample size = 30; solve for = power✓ Δ 6.2×10-14 (13 digits) ↗
Power and sample sizeOne sample talpha = 0.05; analysis = one_sample_t; effect size = 0.5; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizeOne sample talpha = 0.1; analysis = one_sample_t; effect size = 0.5; sample size = 2; solve for = power✓ Δ 9.9×10-14 (12 digits) ↗
Power and sample sizeOne sample talternative = greater; analysis = one_sample_t; effect size = 0.5; sample size = 30; solve for = power✓ Δ 1.6×10-13 (12 digits) ↗
Power and sample sizeOne sample talternative = less; analysis = one_sample_t; effect size = 0.5; sample size = 30; solve for = power✓ Δ 1.6×10-13 (12 digits) ↗
Power and sample sizeOne sample tanalysis = one_sample_t; effect size = -1✓ ↗
Power and sample sizeOne sample tanalysis = one_sample_t; effect size = 0.0001; sample size = 30; solve for = power✓ Δ 5.2×10-15 (12 digits) ↗
Power and sample sizeOne sample tanalysis = one_sample_t; effect size = 0.5; solve for = sample_size; target power = 0.85✓ exact ↗
Power and sample sizeOne sample tanalysis = one_sample_t; effect size = 1e-08; sample size = 30; solve for = power✓ Δ 6.2×10-17 (14 digits) ↗
Power and sample sizeOne sample tanalysis = one_sample_t; sample size = 1; solve for = power✓ ↗
Power and sample sizeOne way ANOVAanalysis = one_way_anova; effect size = 0.5; sample size = 30; solve for = power✓ Δ 1.7×10-10 (9 digits) ↗
Power and sample sizeOne way ANOVAanalysis = one_way_anova; effect size = 0.5; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizePaired tanalysis = paired_t; effect size = 0.5; sample size = 30; solve for = power✓ Δ 6.2×10-14 (13 digits) ↗
Power and sample sizePaired tanalysis = paired_t; effect size = 0.5; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizePoisson ratealternative = greater; exposure = 10; family = poisson_rate; null rate = 0.1; poisson rate = 0.2; sample size = 15; solve for = power✓ Δ 1.1×10-16 (15+ digits) ↗
Power and sample sizeRaw ANOVAfamily = raw_anova; solve for = power; standard deviation = 3✓ Δ 6.0×10-17 (15+ digits) ↗
Power and sample sizeRaw proportiondesign = paired; family = raw_proportion; null proportion = 0.5; proportion = 0.6; solve for = power✓ ↗
Power and sample sizeRaw proportionfamily = raw_proportion; null proportion = 0.5; proportion = 0.6; sample size = 50; solve for = power✓ Δ 1.7×10-16 (15+ digits) ↗
Power and sample sizeRaw tdesign = independent; family = raw_t; mean difference = 1.2; ratio = 1.5; sample size = 20; solve for = power; standard deviation = 2✓ Δ 3.1×10-13 (12 digits) ↗
Power and sample sizeRaw tfamily = raw_t; mean difference = 1; sample size = 30; solve for = power; standard deviation = 2✓ Δ 6.1×10-14 (13 digits) ↗
Power and sample sizeRaw tfamily = raw_t; mean difference = 1; solve for = sample_size; standard deviation = 2✓ Δ 1.7×10-13 (12 digits) ↗
Power and sample sizeTwo proportionanalysis = two_proportion; effect size = 0.5; ratio = 1; sample size = 30; solve for = power✓ Δ 2.8×10-16 (15+ digits) ↗
Power and sample sizeTwo proportionanalysis = two_proportion; effect size = 0.5; ratio = 1; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizeTwo sample tanalysis = two_sample_t; effect size = 0.5; ratio = 1.5; sample size = 30; solve for = power✓ Δ 2.0×10-13 (12 digits) ↗
Power and sample sizeTwo sample tanalysis = two_sample_t; effect size = 0.5; ratio = 1; solve for = sample_size; target power = 0.8✓ exact ↗
Power and sample sizeUnsupportedanalysis = unsupported✓ ↗
Power and sample sizeZfamily = z; known sigma = 2; mean difference = 1; sample size = 25; solve for = power✓ Δ 1.1×10-16 (15+ digits) ↗
Power for two-level designsFullalpha = 0.05; delta = 1; replicates = 2; sigma = 1; solve for = power; target power = 0.8✓ Δ 5.2×10-13 (11 digits) ↗
Power for two-level designsFullalpha = 0.05; delta = 1; replicates = 2; sigma = 1; solve for = replicates; target power = 0.8✓ Δ 7.3×10-14 (13 digits) ↗
Power for two-level designsFullalpha = 0.1; delta = 1; replicates = 2; sigma = 1; solve for = power; target power = 0.8✓ Δ 2.1×10-13 (12 digits) ↗
Power for two-level designsMainalpha = 0.05; delta = 1; replicates = 2; sigma = 1; solve for = power; target power = 0.8✓ Δ 8.9×10-13 (11 digits) ↗
Reliability demonstration and estimation plansDuration for ten units with an exponential modelconfidence = 0.95; mission time = 1000; mode = duration; sample size = 10; search dimension = test_duration; target reliability = 0.9✓ Δ 3.2×10-10 (12 digits) ↗
Reliability demonstration and estimation plansFailure-terminated exponential chi-square ruleallowed failures = 1; confidence = 0.95; mission time = 1000; mode = duration; sample size = 10; scheme = failure_terminated; target reliability = 0.9✓ Δ 2.9×10-13 (15+ digits) ↗
Reliability demonstration and estimation plansOne allowed failure at 90% confidenceallowed failures = 1; confidence = 0.9; mission time = 1000; mode = binomial; target reliability = 0.95✓ exact ↗
Reliability demonstration and estimation plansWeibullconfidence = 0.95; distribution = weibull; mission time = 1000; mode = duration; sample size = 10; search dimension = test_duration; target reliability = 0.9✓ ↗
Reliability demonstration and estimation plansZero-failure binomial demonstrationconfidence = 0.95; mission time = 1000; mode = binomial; target reliability = 0.9✓ Δ 1.1×10-16 (14 digits) ↗
Sample size for precision and toleranceDistribution-free full-range coverageconfidence = 0.95; content = 0.9; criterion = coverage; max n = 500; mode = distribution_free_tolerance✓ Δ 3.9×10-18 (15+ digits) ↗
Sample size for precision and toleranceHowe two-sided normal tolerance factorassumed sd = 2.4; confidence = 0.95; content = 0.9; criterion = factor; max n = 500; mode = normal_tolerance; sided = two_sided; target = 2.5✓ Δ 1.4×10-15 (15+ digits) ↗
Sample size for precision and toleranceStudent-t mean half-widthassumed sd = 2.4; confidence = 0.95; criterion = absolute; max n = 500; mode = mean; target = 0.75✓ Δ 2.2×10-16 (15+ digits) ↗
Sample size for precision and toleranceWilson proportion half-widthassumed proportion = 0.35; confidence = 0.95; criterion = absolute; max n = 500; mode = proportion; target = 0.1✓ Δ 1.4×10-17 (15+ digits) ↗

Design › Simulation

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Probability calculatorA different recorded fingerprint is reported, not accepteddistribution = family normal, mean 5, standard deviation 2; recorded = no; rows = 50; seed = 12345✓ ↗
Probability calculatorA negative standard deviation is refuseddistribution = family normal, standard deviation -1; x values = 0✓ ↗
Probability calculatorA quantile probability of 1 is refuseddistribution = family normal; probabilities = 1✓ ↗
Probability calculatorAn empty request is refuseddistribution = family poisson, mean 2✓ ↗
Probability calculatorBinomial with p = 0 is degeneratedistribution = family binomial, trials 5, probability 0; x values = 0, 1✓ exact ↗
Probability calculatorBinomial(10, 1/2): P(2 < X <= 5)distribution = family binomial, trials 10, probability 0.5; interval = lower 2, upper 5, lower inclusive no✓ exact ↗
Probability calculatorBinomial(10, 1/2): P(2 <= X < 5)distribution = family binomial, trials 10, probability 0.5; interval = lower 2, upper 5, upper inclusive no✓ Δ 6.0×10-17 (15+ digits) ↗
Probability calculatorColumn names that differ only in case are refuseddistribution = family normal; rows = 5; seed = 1✓ ↗
Probability calculatorMore than 1,000,000 cells is refuseddistribution = family normal; rows = 500001; seed = 1✓ ↗
Probability calculatorOne row is the smallest tabledistribution = family exponential, scale 2; rows = 1; seed = 0✓ exact ↗
Probability calculatorPoisson(3.7) values and discrete quantilesdistribution = family poisson, mean 3.7; probabilities = 0.05, 0.5, 0.95; x values = 0, 2, 2.5, 9✓ Δ 5.3×10-17 (14 digits) ↗
Probability calculatorProbability calculator: Fdistribution = family f, numerator df 3, denominator df 12; interval = lower 0.5, upper 4; probabilities = 0.1, 0.5, 0.99; x values = 0.25, 1, 3.49, 9✓ Δ 1.7×10-15 (14 digits) ↗
Probability calculatorProbability calculator: chi squaredistribution = family chi_square, degrees of freedom 7; interval = lower 2, upper 12; probabilities = 0.05, 0.5, 0.95; x values = 0.5, 7, 14, 30✓ Δ 1.8×10-15 (14 digits) ↗
Probability calculatorProbability calculator: exponentialdistribution = family exponential, scale 2.5; interval = lower 1, upper 4; probabilities = 0.05, 0.5, 0.9; x values = 0.1, 1, 2.5, 12✓ Δ 1.8×10-16 (15+ digits) ↗
Probability calculatorProbability calculator: normaldistribution = family normal, mean 10, standard deviation 2; interval = lower 8, upper 12; probabilities = 0.025, 0.5, 0.95; x values = 7, 10, 13.5, 20✓ Δ 1.9×10-15 (14 digits) ↗
Probability calculatorProbability calculator: tdistribution = family t, degrees of freedom 4.5; interval = lower -1, upper 2; probabilities = 0.01, 0.5, 0.975; x values = -3, 0, 1.25, 6✓ Δ 2.7×10-15 (15+ digits) ↗
Probability calculatorProbability calculator: weibulldistribution = family weibull, shape 1.7, scale 3; interval = lower 1, upper 5; probabilities = 0.1, 0.632, 0.99; x values = 0.2, 1.5, 3, 9✓ Δ 9.0×10-16 (15+ digits) ↗
Probability calculatorQuantile at a cumulative step returns that stepdistribution = family binomial, trials 4, probability 0.5; probabilities = 0.3125✓ Δ 6.0×10-17 (15+ digits) ↗
Probability calculatorReplaying the recorded operation recreates the tabledistribution = family normal, mean 5, standard deviation 2; recorded = yes; rows = 50; seed = 12345✓ exact ↗
Probability calculatorSeeded Poisson table uses integer columnsdistribution = family poisson, mean 3.5; rows = 20; seed = 7✓ Δ 3.5×10-17 (15+ digits) ↗
Probability calculatorSeeded normal table: shape, names and theoretical momentsdistribution = family normal, mean 5, standard deviation 2; rows = 50; seed = 12345✓ exact ↗
Probability calculatorStudent t with one degree of freedom has no meandistribution = family t, degrees of freedom 1; x values = 1✓ Δ 2.2×10-16 (15+ digits) ↗
Probability calculatorUpper tail below the double range keeps its logarithmdistribution = family normal, mean 0, standard deviation 1; x values = 40✓ Δ 5.6×10-14 (15+ digits) ↗
Simulate modelSIMULATION-FIXED 1confidence level = 0.95; draws = 100; inputs = name C, source fixed, value 1; model kind = stored; percentiles = 25, 50, 75; stored model = family ols, response Response, predictors name C, kind numeric, feature order Intercept, C, coefficient names Interce...; thresholds = 11, 12, 13✓ Δ 4.0×10-15 (14 digits) ↗
Simulate modelSIMULATION-FIXED 2confidence level = 0.95; draws = 200; inputs = name C, source fixed, value -1; model kind = stored; percentiles = 25, 50, 75; stored model = family ols, response Response, predictors name C, kind numeric, feature order Intercept, C, coefficient names Interce...; thresholds = 7, 8, 9✓ Δ 3.6×10-15 (15+ digits) ↗
Simulate modelSIMULATION-FIXED 3confidence level = 0.9; draws = 100; inputs = name C, source fixed, value 1; model kind = stored; percentiles = 25, 50, 75; stored model = family ols, response Response, predictors name C, kind numeric, feature order Intercept, C, coefficient names Interce...; thresholds = 11, 12, 13✓ Δ 4.0×10-15 (15+ digits) ↗
Simulate modelSIMULATION-FIXED 4draws = 100; inputs = name C, source entered, distribution family normal, mean 0, standard deviation 1; model kind = stored; stored model = family ols, response Response, predictors name C, kind numeric, feature order Intercept, C, coefficient names Interce...✓ exact ↗

Predict

WindowAnalysisSettings comparedNISTRAnalyticalPublished exampleIndependent reviewer
Fit and score regression modelsA target without a fitted predictor is refusedstored fit supplied✓ ↗
Fit and score regression modelsAn unseen categorical level is rejectedstored fit supplied✓ ↗
Fit and score regression modelsLogisticevent value = yes; family = logistic; interval kind = mean✓ exact ↗
Fit and score regression modelsOLSconfidence level = 0.9; family = ols; interval kind = mean✓ Δ 6.6×10-16 (15+ digits) ↗
Fit and score regression modelsOLSfamily = ols✓ Δ 1.8×10-15 (14 digits) ↗
Fit and score regression modelsOLSfamily = ols; interval kind = individual✓ Δ 6.6×10-16 (15+ digits) ↗
Fit and score regression modelsOLSfamily = ols; interval kind = mean✓ Δ 2.9×10-11 (10 digits) ↗
Fit and score regression modelsPoissonfamily = poisson; interval kind = mean✓ exact ↗
Fit and score regression modelsRepeated score column names are refusedstored fit supplied; names = pred, pred✓ ↗
Fit and score regression modelsScore a target with a frozen OLS artifact (individual)stored fit supplied✓ exact ↗
Fit and score regression modelsScore a target with a frozen OLS artifact (mean)stored fit supplied✓ exact ↗
Fit and score regression modelsScore a target with a frozen Poisson artifact and exposurestored fit supplied✓ exact ↗
Fit and score regression modelsScore a target with a frozen WLS artifactstored fit supplied✓ exact ↗
Fit and score regression modelsScore a target with a frozen logistic artifactstored fit supplied✓ exact ↗
Fit and score regression modelsScore columns from the frozen OLS artifactstored fit supplied; names = pred, lo, hi✓ Δ 8.9×10-15 (14 digits) ↗
Fit and score regression modelsWLSfamily = wls; interval kind = individual✓ ↗
Fit and score regression modelsWLSfamily = wls; interval kind = mean✓ Δ 6.5×10-16 (15+ digits) ↗
Model profiler and desirabilityA target on a bound is refusedgoals = response name y, goal target, weight 1, low 4, high 12, target 12; stored fits supplied✓ ↗
Model profiler and desirabilityAn unattainable goal gives joint desirability zerogoals = response name y, goal maximize, weight 1, low 100, high 200; stored fits supplied✓ exact ↗
Model profiler and desirabilityOLS profile, desirability and the best cornergoals = response name y, goal maximize, weight 2, low 0, high 20; stored fits supplied✓ Δ 8.7×10-15 (15+ digits) ↗
Model profiler and desirabilityPoisson profile at an entered exposure of 2.5goals = response name cnt, goal minimize, weight 1, low 0, high 60; stored fits supplied✓ Δ 6.4×10-14 (14 digits) ↗
Model profiler and desirabilityPredictors must cover every retained rolegoals = response name y, goal maximize, weight 2, low 0, high 20; stored fits supplied✓ ↗
Model profiler and desirabilityTarget goal at a 90% confidence levelconfidence level = 0.9; goals = response name y, goal target, weight 1, low 4, high 12, target 7; stored fits supplied✓ Δ 5.5×10-15 (14 digits) ↗
Model profiler and desirabilityTwo models with weighted joint desirabilitygoals = response name y, goal maximize, weight 2, low 0, high 20, response name yb, goal maximize, weight 1, low 0, high 1; stored fits supplied✓ Δ 2.4×10-15 (14 digits) ↗
  1. 1 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.0e-13 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  2. 2 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within -1.3e-13 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  3. 3 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.7e-14 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  4. 4 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.1e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  5. 5 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.9e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  6. 6 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.0e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  7. 7 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.2e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  8. 8 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 5.6e-14 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  9. 9 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.1e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  10. 10 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 6.9e-16 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  11. 11 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 5.5e-16 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  12. 12 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.1e-10 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  13. 13 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 4.5e-10 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  14. 14 the fitted SSE is strictly below the certified SSE (HypoNum found the better optimum in a flat valley), so the digit shortfall against the certified parameters is the degenerate objective, not a solver defect.
  15. 15 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.5e-13 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  16. 16 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.0e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  17. 17 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.6e-11 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  18. 18 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 4.9e-11 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  19. 19 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 3.7e-14 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  20. 20 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 3.6e-13 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  21. 21 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.4e-13 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  22. 22 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.7e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  23. 23 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.3e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  24. 24 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 6.7e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  25. 25 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 6.5e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  26. 26 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 3.6e-12 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  27. 27 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within -2.5e-14 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  28. 28 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 6.6e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  29. 29 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 2.0e-14 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  30. 30 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 9.3e-15 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  31. 31 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.1e-10 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  32. 32 optimum confirmed by independent arithmetic - the fitted SSE matches the certified SSE within 1.3e-10 relative (gradient diagnostics at both points sit at the same finite-difference noise floor), so the coefficient/covariance digit shortfall is the flat, ill-conditioned objective, not a stopping error.
  33. 33 This reference dataset is deliberately ill-conditioned. The NIST certified values were computed in extended precision; any double-precision solver, including an independent SVD least-squares fit, agrees to about 12 to 13 significant digits. HypoNum's agreement is shown as observed.
  34. 34 This reference design has a condition number above 10¹³. In double precision the raw-basis coefficients and standard errors cannot be reproduced to the certified digits by any solver; only extended-precision arithmetic can. The digits shown are what double precision achieves.

Not validated yet: 16

  • 4 analysis suite has no comparison case yet
  • 0 model not available in HypoNum
  • 12 reference published to too few digits to assess
  • 0 comparison not passed on this build
SurveyCalibrate weightsanalysis suite has no comparison case yet
SurveyCount regressionanalysis suite has no comparison case yet
SurveyGenerate replicate weightsanalysis suite has no comparison case yet
SurveyMultinomial regressionanalysis suite has no comparison case yet
Quality › Measurement systemsGage R repeatability screeningANOVAreference published to too few digits to assess
Quality › Measurement systemsGage R repeatability screeningAverage rangereference published to too few digits to assess
Quality › Measurement systemsGage R repeatability screeningChart worksheetreference published to too few digits to assess
Quality › Measurement systemsGage R repeatability screeningPooled SDreference published to too few digits to assess
Quality › Measurement systemsGage R repeatability screeningRangereference published to too few digits to assess
Quality › Measurement systemsGage R&RANOVAreference published to too few digits to assess
Quality › Measurement systemsGage R&RAverage rangereference published to too few digits to assess
Quality › Measurement systemsGage R&RChart worksheetreference published to too few digits to assess
Quality › Measurement systemsGage R&RPooled SDreference published to too few digits to assess
Quality › Measurement systemsGage R&RRangereference published to too few digits to assess
Quality › Reliability engineeringDegradation-based lifetimeDegradation-based lifetime: inverse-kelvin temperature acceleration to a use temperaturereference published to too few digits to assess
Quality › Reliability engineeringDegradation-based lifetimeDegradation-based lifetime: repeated measurements, lognormal rates, no temperature axisreference published to too few digits to assess
Methods

Accuracy policy. A comparison passes when:

Quantity Requirement
Counts, labels and decisions exact equality
Probabilities and dispersion statistics (variances, standard errors, standard deviations) relative error at most 10⁻⁴
Other quantities whose reference is at most 10⁻⁹ in magnitude absolute error at most 10⁻⁹
Other quantities relative error at most 10⁻⁴
Stated mathematical or statistical conditions the condition holds

Many comparisons agree far more closely than the policy requires; the matrix shows the agreement actually observed.

Digits agreed. For a reference c and an error E, the agreement is −log₁₀(E / |c|), rounded down to a whole number and capped at 15. “Exact” means E = 0. Where a reference is published to fewer digits than the product computes, the comparison is listed as not assessed rather than counted either way.

Not assessed. One reason per group:

Cases Reason
Gehan survival examples (2) the published tables print the statistics to three or four digits, too few to score
AIAG measurement-system examples (5) the MSA manual prints its worked results to four or five digits
Dunnett critical value (1) the reference comes from another software library without an independent error bound; a quadrature reference is being prepared
Degradation lifetime examples (2) the published reference is an approximation with no stated error bound