statsmodels · k-dense-ai — Claude Code skill: regressions and econometrics

What statsmodels does

The skill covers statsmodels in Python: linear models (OLS, WLS, GLS, quantile regression), generalized linear models, discrete choice, count and ordinal outcomes, ARIMA, SARIMAX, VAR and forecasting. Diagnostics get their own treatment — heteroskedasticity, autocorrelation, residual normality, outliers and influential points, model comparison via AIC/BIC and likelihood ratio tests.

Inside are reference files: a quick start with small worked examples for OLS, logistic regression, ARIMA and GLM, an overview of modeling capabilities, model selection and the R-style formula API, plus per-topic files on linear models, GLM, discrete choice, time series and diagnostics. The agent gets Read, Write, Edit and Bash.

You need Python 3.10+ and statsmodels 0.15.0; the tested NumPy 2.5.3 and SciPy 1.18.1 stack calls for Python 3.12+. Plotting requires matplotlib, predictive metrics require scikit-learn. Network access matters only for installation or docs, and no credentials are involved. Install into a dedicated environment: uv pip install statsmodels==0.15.0 numpy==2.5.3 scipy==1.18.1 pandas==3.0.6 matplotlib==3.11.2 scikit-learn==1.9.1. Mismatched NumPy and SciPy versions are where things break — 0.15 moved several tests to result_object=True with named fields, and rng= replaces the older seed and random_state arguments.