A Minitab-style statistics workbench for Windows — a column worksheet, a Minitab-style
Session output pane, and analyses run from a Stat/Graph menu. Built with .NET 10 +
WPF, Math.NET Numerics (math) and ScottPlot (graphs).
- Worksheet — editable column grid (C1, C2, …); import/export CSV/TSV and Excel (.xlsx);
native
.ssprojproject files; missing values as*. File ▸ Sample Data loads ten built-in example datasets tuned to the various analyses. - Basic Statistics — descriptive statistics (Minitab quartile method), 1-sample / 2-sample (pooled & Welch) / paired t, 1- & 2-proportion, chi-square goodness-of-fit and association, Pearson/Spearman correlation, Anderson-Darling normality, F-test for two variances, Fisher's exact.
- Nonparametrics — Mann-Whitney, Wilcoxon signed-rank, Kruskal-Wallis, sign test, runs test.
- ANOVA — one-way (with Tukey HSD post-hoc), two-way (with interaction), tests for equal variances (Bartlett, Levene).
- Regression — simple, multiple, polynomial, best-subsets (Mallows Cp), stepwise, and binary logistic; coefficient SE/t/p, R²/adj-R²/S, ANOVA table, fitted-line and residual plots.
- Time series — trend analysis, moving average, single/double/Winters exponential smoothing, classical decomposition, ACF/PACF, ARIMA(p,d,q) and seasonal SARIMA(p,d,q)(P,D,Q)ₛ (forecasts with confidence bands); MAPE/MAD/MSD accuracy and actual/fitted/forecast plots.
- Multivariate — principal components (scree plot), factor analysis (varimax rotation), k-means clustering.
- Bayesian — Beta-Binomial proportion, 1-sample normal mean (known & unknown variance), and reference-prior linear regression: posterior summaries, credible intervals, tail probabilities.
- Mixed / hierarchical — one-way random-effects (random-intercept) model: variance components, intraclass correlation, and shrinkage (BLUP) group estimates.
- Reliability / survival — parametric life-data fitting (Weibull/exponential/lognormal/normal MLE + Weibull probability plot) and Kaplan-Meier survival. Both accept an optional 0/1 right-censoring column; censored units contribute survival time to the likelihood, and the probability plot uses Johnson rank-adjusted plotting positions. Parameters, percentiles and the mean/StDev/median all carry standard errors and confidence intervals — observed Fisher information for the parameters, the delta method for functions of them.
- DOE — create full 2^k / 2^(k-p) fractional factorial designs (generators, resolution, defining relation), response-surface designs (central composite + Box-Behnken), and mixture designs (simplex-lattice + simplex-centroid); analyze 2-level factorials (effects, Pareto), second-order response surfaces, and Scheffé mixture models.
- Calculator — compute a new column from a formula:
Cn/named columns,+ - * / ^, per-row functions (sqrt, log, exp, trig, round) and aggregates (mean, sum, stdev, …). - Control charts (SPC) — Xbar-R, Xbar-S, I-MR, P, NP, C, U (with Nelson tests 1 & 2).
- Quality tools — normal process capability (Cp, Cpk, Pp, Ppk, Cpm); crossed Gage R&R (ANOVA method: variance components, %study var, distinct categories); power & sample-size calculators (1-/2-sample t, 1-proportion).
- Graphs — histogram, boxplot, scatterplot, time-series plot, normal probability plot.
StatStudio.slnx
src/StatStudio.Core (net10.0) pure, UI-free engines — Data/, Statistics/, Statistics/Spc/, Inference/
src/StatStudio.Wpf (net10.0-windows) WPF UI, dialogs, ScottPlot graphs (x64)
tools/StatStudio.Smoke(net10.0) deterministic numeric self-tests
assets/ make-icon.ps1, capture.ps1
installer/ StatStudio.iss, build-installer.ps1
The editable DataTable is the live worksheet; analyses use a detached Core Worksheet
snapshot after committing edits. Declared column types survive this conversion. Numeric results
are stored at full precision and rounded only for display. Column-header sorting is disabled so
the visible worksheet, ordered analyses, and exports keep the same observation order.
Imports let you choose whether the first row contains column names. For a CSV exported by
StatStudio, select the export option to decode its reversible formula guards; leave it off for
external CSVs to preserve literal apostrophes. XLSX uses text/quote-prefix metadata instead.
Use .ssproj to preserve declared text identifiers such as 00123 and other column types.
Time-series, I-MR, and capability inputs reject missing observations before the final value; trailing empty worksheet padding is ignored. Moving averages use trailing windows for both odd and even lengths. Power calculations honor signed alternatives and return whole sample sizes.
Heavy analyses and file operations run in the background with a Cancel operation button. Numerical loops check cancellation; third-party file/matrix calls finish their current call before cancellation can be observed. Limits are 10,000 forecasts/design runs, 512 worksheet columns, 2,000,000 worksheet cells, and 64 MB per imported file.
dotnet build StatStudio.slnx
dotnet run --project src/StatStudio.Wpf
dotnet test StatStudio.slnx
dotnet run --project tools/StatStudio.Smoke # optional verbose smoke output
StatStudio.Wpf --demo loads a sample dataset and runs a few analyses (see ProcessArgs for the
--shot-* screenshot flags used during development).
powershell -ExecutionPolicy Bypass -File installer/build-installer.ps1 # -> dist/StatStudioSetup.exe
powershell -ExecutionPolicy Bypass -File tools/e2e-install.ps1 -InstallDir C:\temp\StatStudio-e2e
Self-contained win-x64 (no .NET prerequisite); installs machine-wide under Program Files and requests administrator approval.
StatStudio covers Minitab's mainstream surface and beyond: basic statistics, nonparametrics, ANOVA/regression (incl. logistic), the full DOE suite (factorial, fractional, response-surface, mixture), SPC + capability + Gage R&R, the full time-series suite (smoothing, decomposition, ARIMA/SARIMA), multivariate (PCA, factor analysis, clustering), reliability/survival, power & sample size, conjugate Bayesian inference, one-way mixed/hierarchical models, and a worksheet calculator — all engine-verified by the discoverable smoke test suite. The engine-per-analysis + dialog + Session-output design keeps further additions incremental.