Linear regression

Skill WindcleaverDev/regkit/.agents/skills/linear-regression

Statistically rigorous regression skills for Claude — Python computes, the model narrates.

Install

npx -y skills add WindcleaverDev/regkit --skill linear-regression

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Fit a linear regression (OLS) on tabular data and produce a rigorous report — coefficient table with confidence intervals, standardized betas, plain-English interpretation of each coefficient (transform-aware), fit statistics, and a standalone HTML report. Use this skill whenever the user wants to model a continuous outcome from one or more predictors, asks to "fit a regression" or "run OLS" or "model Y from X", or hands over tabular data with a continuous target. Outputs are both a structured JSON (LinearRegressionReport) and a self-contained HTML deliverable. Pair this skill with the diagnostics skill to check assumptions and identify influential observations.

SKILL.md

2.9 KB, as published. Nobody here has run it

linear-regression

Fits an OLS linear regression with statsmodels, produces a LinearRegressionReport JSON validating against the pack schema, and renders a standalone HTML report.

When this skill fires

User wants to fit a linear regression on tabular data
User has identified a continuous target and one or more predictors
User asks for OLS, multiple regression, or "model Y from X"
A previous skill (e.g. pre-analysis) recommended linear regression

Inputs

A CSV or Parquet file with the data
The target column name (must be numeric)
A list of predictor column names (numeric or categorical — categoricals are one-hot encoded with the first level dropped)

Optional:

--log-target to fit on log(target) — useful for skewed positive targets
--robust-se {HC0|HC1|HC2|HC3} for heteroscedasticity-robust standard errors
--standardize to also report standardized β coefficients

How to invoke

uv run python linear-regression/scripts/fit.py \
    --data path/to/data.csv \
    --target price \
    --features sqft,bedrooms,bathrooms,neighborhood \
    --output results/

Outputs results/report.json (LinearRegressionReport) and results/report.html.

Verbalising the output

The interpretations field contains a list of InterpretationFact objects. Each has:

fact — the canonical claim
confidence — high/medium/low based on p-value and CI width
caveats — list of qualifiers ("ceteris paribus", "not causal", scale notes)

Read the headline field aloud first, then walk through the top 2-3 interpretation facts ordered by absolute coefficient size, attaching caveats appropriate to the user's apparent sophistication. Do not invent statistics; only verbalise what's in the report.

Diagnostics

This skill performs the fit — it does not run the full assumption battery. After fitting, recommend the user run the diagnostics skill on the fitted model:

uv run python diagnostics/scripts/diagnose.py --fit-report results/report.json --data path/to/data.csv --output results/

Reference files

references/interpretation.md — when each interpretation_type applies and how to phrase it
references/robust_se.md — when to use each HC variant

Gives 0 of the 12 instructions most data analysis skills give

Counted across 286 of the 286 authors here whose files we hold, read 2026-08-06

use excel formulas instead of hardcoded calculated valuesin 35 of 286, across 7 files
match existing template conventions when modifying filesin 35 of 286, across 7 files
document sources for all hardcoded valuesin 35 of 286, across 7 files
write minimal concise python codein 35 of 286, across 7 files
place all assumptions in separate assumption cellsin 32 of 286, across 5 files
apply industry-standard color coding to financial modelsin 31 of 286, across 5 files
format years as text stringsin 30 of 286, across 3 files
recalculate formulas using recalc.py after modificationsin 30 of 286, across 3 files
format negative numbers using parenthesesin 30 of 286, across 3 files
fix all identified formula errors before finishingin 27 of 286, across 1 file
use colorblind-safe palettesin 19 of 286, across 12 files
Name tests after the prevented bugin 13 of 286, across 8 files

Said here and by no other author read

drop the first level when one-hot encoding categoricals
read the headline field aloud first
verbalise only top facts by coefficient size
attach caveats to interpretation facts
recommend running the diagnostics skill after fitting

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.