MarginalEffects (Model to Meaning)

What it does

This skill encapsulates the manual and pedagogical material for the marginaleffects package and the companion book ‘Model to Meaning’. It helps agents guide users through choosing estimands (predictions, comparisons, slopes), constructing counterfactual grids, aggregating results (ATE, ATT, CATE), and selecting inference methods (delta, bootstrap, Bayesian). The skill includes language-specific examples for R and Python and explains how to use core functions like predictions(), comparisons(), slopes(), avg_* variants, and datagrid().

When to use it

Use when users ask how to interpret model outputs, compute marginal effects or average treatment effects, set up counterfactual comparisons, or run hypothesis tests on derived quantities. It's appropriate for both conceptual framing (five-question framework) and concrete code examples in R or Python.

What's included

• Scripts: none bundled (has_scripts=false)
• References: links and book chapters summarized in the skill body (has_references=false)
• Instructions: five-question framework (Quantity, Grid, Aggregation, Uncertainty, Test), function reference pointers, and example code for logit models in R and Python.

Compatible agents

Agents capable of generating or reviewing R/Python statistical code (Codex/Copilot-style assistants, Claude Code, Cursor) are the best fit; also useful for agents that help interpret outputs to non-technical stakeholders.