StatLab covers the everyday procedures of a first statistics course — descriptives, crosstabs and chi-square, t tests, one-way ANOVA, correlation, linear regression, reliability and the common nonparametric tests — with 88 features live today. 57 of them are validated number-for-number against scipy and statsmodels (433 reference test cases, 36,554 values checked against scipy/statsmodels/numpy, 0 failures).
The table lists the capabilities people usually look for in a desktop statistics package and StatLab's status for each. Live = available now; planned = on the roadmap (wave shown); parked = not scheduled yet. “Validated” means every number was checked against independent reference software to a relative tolerance of one in a million. StatLab is an independent product and is not affiliated with IBM; SPSS is a trademark of IBM Corp.
Data editor
Feature
Status
Validated
Data View grid (edit, navigate, value-label display)
SPSS-familiar dialogs: variable list + arrow buttons, Statistics/Options/Plots/Post Hoc sub-dialogs, Paste to syntax, Reset
Live
Tested in the app
Analyze menu tree with sub-menus (Descriptive Statistics, Compare Means, General Linear Model, Correlate, Regression, Classify, Dimension Reduction, Scale, Nonparametric Tests, Survival)
Live
Tested in the app
Import / export
Feature
Status
Validated
CSV / TSV import (header row, type inference)
Live
Tested in the app
Excel .xlsx/.xls/.ods import (sheet pick)
Live
Tested in the app
SPSS .sav import (labels, value labels, missing, measure, long names, very long strings, UTF-8; bytecode compression)
Live
Yes — reference-checked
Text import options: delimiter (auto/comma/semicolon/tab/pipe), decimal comma, encoding (UTF-8/Windows-1252/ISO-8859-1/UTF-16), header row, skipped lines, preview; Excel sheet + range; paste data from Excel
Live
Yes — reference-checked
Compressed .zsav import
Live
Yes — reference-checked
SPSS portable .por import
Live
Yes — reference-checked
Stata .dta import (releases 117/118/119 incl. strL, value labels) and export (118); cross-checked with pandas both ways
Live
Yes — reference-checked
SAS .sas7bdat import (32/64-bit, both byte orders, RLE and RDC compression, labels, encodings, dates as ISO text); matches pandas on 28 sample files (29,431 values)
Other post hocs (Bonferroni, Scheffé, Games-Howell, Dunnett), contrasts, Brown-Forsythe
Planned · wave 2
General linear model
Feature
Status
Validated
GLM Univariate: factorial ANOVA and ANCOVA, Type I/II/III/IV sums of squares, WLS weights, empty-cell designs, random factors with expected mean squares, estimated marginal means, profile plots
Live
Yes — reference-checked
GLM contrasts and follow-ups: custom contrasts, simple effects, homogeneous subsets (Tukey, S-N-K, REGWQ), post hoc Tukey/Bonferroni/Scheffé/Games-Howell/Dunnett/Tamhane T2, Levene (mean, median with adjusted df)
Rotated factor solutions are the same as SPSS's up to sign and order: StatLab reflects each factor so its loadings sum positive and orders factors by sum of squared loadings, so some factors may print with the opposite sign or in a different order. Promax, equamax, K-means start-point choice and the Hosmer-Lemeshow grouping follow the published definitions and are formula-checked, not compared with SPSS output. GLM, regression, logistic, factor, cluster and discriminant numbers are checked against statsmodels/scipy/scikit-learn references (95 cases, 14,547 values, 0 failures).
Discriminant: Mahalanobis / Rao's V selection, separate-groups covariance, territorial map
Planned · wave 4
Nonparametric
The add-on nonparametric, survival, ROC, time-series, effect-size, power, missing-data and reliability procedures are checked against scipy, statsmodels, lifelines and pingouin on reference data (105 cases, 1,847 values, 0 failures) and by exhaustive enumeration for exact tests; not yet compared with SPSS output itself.
Feature
Status
Validated
Mann-Whitney U, Wilcoxon signed-rank, Kruskal-Wallis H (tie-corrected)
Live
Yes — reference-checked
Legacy dialogs, full set: chi-square goodness of fit, binomial, runs, one-sample K-S, 2 independent (Mann-Whitney + exact p), 2 related (Wilcoxon, sign, McNemar, marginal homogeneity), K independent (Kruskal-Wallis, median, Jonckheere-Terpstra), K related (Friedman, Kendall's W, Cochran's Q)
Live
Yes — reference-checked
Exact or Monte Carlo p-values for Kruskal-Wallis, Friedman, Jonckheere-Terpstra and k-sample median tests
Power and sample size: t tests, one-way ANOVA, regression, correlation, proportions, chi-square
Live
Yes — reference-checked
Charts
Limits, stated honestly: PDF export is vector everywhere (standard Helvetica text; clip paths and gradients are not drawn); Ward's method uses the scipy (Euclidean) scale, which differs from SPSS's coefficient scale; ROC significance uses the DeLong (nonparametric) standard error, whereas SPSS uses the standard error under the null (AUC and its CI are unaffected); dendrograms are limited to 4,000 cases and scatterplot matrices to 8 variables. Chart numbers (bins, quartiles, whiskers, fits, LOESS, CIs, KM, ROC, KDE, eigenvalues, linkage) are checked against numpy/scipy/statsmodels: 135 cases, 18,889 values, 0 failures.
Feature
Status
Validated
Histogram (+ normal curve), bar, boxplot with outliers/extremes, scatter + fit (original Graphs dialogs)
Live
Yes — reference-checked
Chart Builder: drag or select variables onto X / Y / colour / panel; 24 chart types; templates (style only)
Live
Yes — reference-checked
Bar: simple, clustered, stacked, 100%, horizontal, error bars (95% CI, ±SE, ±SD); line, area; pie/donut; dual axis; population pyramid
Live
Yes — reference-checked
Scatter: grouped, linear/quadratic/cubic fit with R², mean and individual CI bands, LOESS; scatterplot matrix
Live
Yes — reference-checked
Normal Q-Q and P-P (plain and detrended); regression residual plots (ZRESID × ZPRED, histogram, P-P, leverage with Cook's D)
Interactive charts: hover values, zoom/pan, click a point to select the case in Data View; keyboard navigation
Live
Tested in the app
APA chart theme by default; light/dark; Okabe-Ito, viridis, greyscale palettes; journal size presets; PNG 300/600 dpi, SVG, PDF
Live
Tested in the app
Alt text written from the data for every chart
Live
Tested in the app
Charts exported into Word (.docx) output
Planned · wave 2
Vector PDF for every chart (Output and Chart Builder; selectable text)
Live
Tested in the app
Row and column panel variables (small multiples)
Live
Yes — reference-checked
Click-on-chart editing (titles, series colours, axes)
Live
Tested in the app
3-D scatter (rotatable); Pareto; control charts (X-bar/R/S, I-MR, p/np, c/u with 13 run rules); sequence charts; high-low/range; dot and drop-line; dendrogram up to 4,000 cases (0–25 rescale)
Live
Yes — reference-checked
3-D bar and surface charts, maps, date axes, CUSUM/EWMA charts, process capability (Cp/Cpk)
Planned · wave 4
Output and syntax
Feature
Status
Validated
Output viewer with outline, APA-style text for every analysis
Live
Tested in the app
Syntax log with re-run and replay-all
Live
Tested in the app
Pivot output tables: swap rows and columns (one click, reversible); full pivot trays with layers planned
User guide (27 topics, one page per procedure, glossary, FAQ, search) with “? Help” in every dialog
Live
Tested in the app
Gallery: 10 case studies, 11 practice datasets with codebooks (CSV, Excel, .sav), 14 one-click recipes; filters by analysis, field, level
Live
Yes — reference-checked
Case-study numbers checked against reference software (44 cases)
Live
Yes — reference-checked
Where is my data processed?
In your browser. StatLab downloads once and then computes everything locally; data files you open are never uploaded. If you create an account and choose Save to my profile, an encrypted copy of that project is stored on our server so you can open it on another device. See the privacy notice.