Regress Desk - notice

The app's agent prompt is derived from the agent skill "statsmodels" (@k-dense-ai/statsmodels) in
the repository k-dense-ai/scientific-agent-skills by K-Dense Inc.
  https://github.com/k-dense-ai/scientific-agent-skills (skills/statsmodels)
The skill's front matter declares the BSD-3-Clause licence. No text of the skill is redistributed
verbatim; the prompt was rewritten for this app.

The in-browser fit (olskit.js) is an independent JavaScript implementation of definitions used by
the statsmodels Python package (0.14.6) and by patsy and scipy as statsmodels calls them: OLS with
patsy treatment coding and term order, the summary() statistics, HC3 robust covariance,
variance_inflation_factor, het_breuschpagan (robust), OLSInfluence (hat diagonal, externally
studentized residuals, Cook's distance), scipy.stats.normaltest (D'Agostino-Pearson) and the
condition number as the ratio of the design matrix's singular values. No statsmodels, patsy or
scipy code is included. statsmodels is BSD-3-Clause, https://github.com/statsmodels/statsmodels.

It was checked against statsmodels 0.14.6 (with scipy 1.18.1) on 1,000 random models: one to four
numeric predictors on scales from 1e-3 to 1e4, text and integer categoricals, np.log, squares,
two-way interactions, missing cells, log outcomes and no-intercept fits. Term names, observation
counts, degrees of freedom and the constant detection were identical in every case; coefficients,
standard errors, p-values, intervals, HC3, VIF, Breusch-Pagan, Omnibus, Jarque-Bera, Durbin-Watson,
the condition number and the influence measures agreed within 2e-5 relative in the worst
(ill-conditioned, condition number above 1e9) cases and within 6e-8 when the condition number was
below 1e6. Where a row has leverage 1, statsmodels' HC3 and Cook's distance divide by zero and print
noise; this page reports them as undefined instead.

If this skill contributed to a publication, K-Dense asks that it be cited:
  Kassis, T., Agarwal, V., He, Y., Patel, D., & Brueckner, A. M. (2026). Scientific Agent Skills:
  A Library of Procedural Knowledge for Research Agents. arXiv:2609.00065.

statsmodels itself: S. Seabold and J. Perktold, "statsmodels: Econometric and statistical modeling
with python," Proceedings of the 9th Python in Science Conference, 2010.

The example data in example.js are illustrative, generated from stated made-up models, not
measured: 64 home sales (price from size, bedrooms, age and neighborhood, with noise that grows
with size), 24 plate-reader standards (absorbance linear in concentration plus a small temperature
term and noise), and 80 customers (churn drawn from a logistic model of tenure, tickets and plan).