Free for Windows and Linux
Open your SPSS, Stata, SAS, Excel or CSV data, run the analysis from a menu, and get tables and charts ready for your paper. Every release is tested against results computed in R and Python.
Opens your files directly
What it does
Descriptive statistics, t tests and ANOVA, regression and GLMs, mixed models, survival, time series, meta-analysis, survey designs, factor analysis and more. Effect sizes are reported next to the tests they belong to.
See what's in the Analyze menu
Before each release, the numbers the app prints are compared with values computed independently in Python and in R, and every NIST reference dataset is run. You can repeat the checks on your own computer from Help › Verify this installation.
Read the validation report
Everything you do in the menus is recorded as a command you can edit and run again. Save the script for next month's data, or export a bundle that a colleague can re-run and check against the original.
How reproducibility works
SPSS, Stata and SAS files keep their variable labels, value labels and missing-value codes.
Change titles, colours and legends in place, and export to SVG, PDF or PNG.
Ask in plain language and get the same tables the menus produce. Use Claude, an OpenAI-compatible service or a local Ollama model.
Drive the same engine from Python with pip install chakatastat, or run a script without opening a window.
Edit data directly, with a Variable View for labels, measurement levels and missing values.
A native engine does the work in the background, so the window stays responsive during long fits.
Validation
These checks run on every release. The results, tolerances and sources are published in full.
353
statistics compared with results computed in Python and in R, across 53 procedure families.
58
NIST reference datasets, every one NIST publishes, scored by digits of agreement with the certified values.
3,264
automated tests of the engine and the app, run before each release.
When a statistic has more than one accepted definition, such as a quantile method or a continuity correction, the choice is written down. Each page of the User Guide says how its results are computed.
Scripting
The menus, the script editor, the Python client and the command line all run the same commands on the same engine. A result you got by clicking can be run again from a script.
// Written for you as you use the menus. Edit and run it again.
descriptives variables=[Age, BMI, SystolicBP]
independent_ttest variables=[SystolicBP], grouping=Hypertensive# The calculations run in the ChakataStat engine, not in Python.
import chakatastat as ig
with ig.open("health_study.ckd") as ds:
print(ds.descriptives(variables=["Age", "BMI", "SystolicBP"]))
ds.independent_ttest(variables=["SystolicBP"], grouping="Hypertensive")
df = ds.to_polars() # continue in Polars# Run a script without opening a window and write a report.
ChakataStat --run analysis.cks health_study.csv --out report.html
# Offer the same tools to another program over MCP.
ChakataStat --mcp health_study.csv// Ask in plain language. The reply shows the real tables and charts.
Compare systolic blood pressure between the hypertensive
and non-hypertensive groups, and show me the effect size.The Analyze menu
Frequencies, descriptives, explore, crosstabs, custom tables, missing values and multiple imputation.
t tests, equivalence tests, one-way ANOVA with post hoc tests, GLM, repeated measures and MANOVA.
Linear, logistic, Poisson, negative binomial, ordinal, multinomial, quantile, robust, regularized, nonlinear and instrumental variables.
Linear mixed models, GLMM, GEE, panel fixed and random effects, and mediation with moderation.
Kaplan–Meier, Cox regression, parametric survival and competing risks.
ARIMA, exponential smoothing, seasonal decomposition, stationarity tests, VAR and Granger causality.
Factor analysis, confirmatory factor analysis, reliability, clustering, discriminant analysis, latent class and item response theory.
ROC curves, Bland–Altman, meta-analysis, propensity scores, survey designs, power analysis and bootstrap intervals.
Nonparametric and Bayesian tests sit beside their classical counterparts.
Screenshots
Taken with the sample dataset that comes with the app. Select a picture to enlarge it.
Free for Windows and Linux. A macOS version is not available yet.
Store versions are signed and update automatically. On Debian and
Ubuntu you can also install the .deb from the
releases page.
The Python client is on PyPI: pip install chakatastat.