What is KernelStats?
KernelStats is a free, open-source statistical analysis platform I built to replace expensive commercial tools like Minitab ($1,700/yr), JMP ($1,785/yr), and SAS ($8,700/yr). It runs entirely in your browser — no installation, no license, no barriers.
I built KernelStats because I believe powerful statistical analysis should be accessible to everyone: students learning their first hypothesis test, researchers on a budget, and organizations that shouldn’t need an enterprise contract to run a t-test.
Visit kernelstats.com — no account required.
Features
Data Quality
- Variable Doctor — 15 automated diagnostics detect coded missing values, numeric-as-text, date columns, outliers, and ID columns. One-click fixes with full undo.
- Support for CSV, Excel, SPSS, Stata, SAS, RDS, and direct paste/URL import
Analysis
- 20+ hypothesis tests with a Test Finder wizard that recommends the right test
- 10 regression families — linear, logistic, probit, multinomial, ordinal, Poisson, negative binomial, zero-inflated, hurdle, and bias-reduced
- Time series & GARCH — ARIMA (auto-selection), GARCH(1,1), GJR-GARCH, EGARCH with volatility plots, news impact curves, and forecasting
- Survival analysis — Kaplan-Meier, Cox PH, hazard ratio forest plots, Schoenfeld tests
- Machine learning pipeline — 7 algorithms with auto-detect classification vs regression, model comparison, ROC curves, variable importance
Output
- Plain-English interpretations for every test result
- Professional reports in Word (.docx) or HTML with customizable sections
- Industry templates — SPC control charts (Cp/Cpk), RFM segmentation, Pareto, heatmaps, volcano plots
How it compares
| Feature | KernelStats | Minitab | JMP | SAS |
|---|---|---|---|---|
| Annual cost | Free | $1,700 | $1,785 | $8,700 |
| Open source | ✓ | |||
| Runs in browser | ✓ | |||
| Variable Doctor | ✓ | |||
| Test Finder wizard | ✓ | |||
| GARCH models | ✓ | ✓ | ||
| ML pipeline | ✓ | ✓ | ✓ | ✓ |
| Plain-English output | ✓ |
Technology
KernelStats is built in R with Shiny, powered by ggplot2, and backed by 132 statistical methods mapped to peer-reviewed R packages. The source code is MIT-licensed and available on GitHub.
Links:
- kernelstats.com — Live app
- GitHub — Source code
Coming soon:
- KernelStats Pro — AI-powered version with Claude integration, automated interpretation, and enterprise deployment