statlab

A large, dependency-free statistics & modeling library for JavaScript/TypeScript. 84 method modules + 6 numerical primitives — regression, ANOVA, survival, econometrics, multivariate, Bayesian, nonparametric, time series, and much more. Pure ESM, runtime-agnostic (runs unchanged in Node and the browser), ships ESM + CJS.

Install

npm install statlab

Zero runtime dependencies.

Usage

The public API is namespaced by module (this resolves function-name collisions across modules — e.g. bootstrapCI, hausmanTest, egarch each appear in more than one module).

// Namespaced barrel
import { survival, anova, core } from 'statlab';

survival.coxPH(/* … */);
anova.oneWayANOVA(/* … */);
core.avg([1, 2, 3]);

// Or import a single module directly (tree-shakeable subpath)
import { coxPH } from 'statlab/methods/survival';
import { avg } from 'statlab/math/core';

CommonJS works too:

const { survival } = require('statlab');

Module layout

• statlab/math/* — numerical primitives: core, distributions, inference, matrix, power, rng.
• statlab/methods/* — the 84 statistical method modules (anova, regression, survival, ecology, econometric, sem, bayesian, …).

Note: two modules share the basename power. In the barrel, the low-level primitives are power (statlab/math/power) and the power-analysis methods are powerAnalysis (statlab/methods/power).

Accuracy & rigor

Every headline method is covered by tests, many against independent numeric oracles (R / scipy / statsmodels). Some methods in the long tail use documented approximations (normal-approx criticals, one-term asymptotic p-values, etc.) — see the repository's BASELINE.md for the per-method rigor classification.

Types

v0.x ships auto-generated .d.ts declarations; parameter types are currently loose (any). Precise JSDoc-derived types are being added module-by-module.

License

MIT