marginaleffects

Compute marginal effects by translating model outputs into predictions, contrasts, and slopes using functions like avg_predictions and avg_comparisons. This Skill supports both unit-level and aggregate summaries for R and Python users.

Average predictions provide aggregate marginal estimates across a population, while unit-level slopes offer conditional estimates for individual observations. Both estimands support hypothesis tests for various queries like ATE, ATT, and CATE.

Yes, you can specify uncertainty and use datagrid-like operations to explore predictor spaces and counterfactual scenarios. This allows for detailed by-group analyses and testing across custom predictor grids.

Yes, the framework supports a wide range of estimands including risk ratios, odds ratios, ATE, ATT, and CATE. It applies these computations across diverse model types for both R and Python environments.

The five-question framework is an enforced methodology for clarifying model outputs through marginal effects. It structures the computation of predictions, contrasts, slopes, and hypothesis tests to ensure actionable insights.

Use avg_comparisons when standard model outputs are opaque and you need actionable marginal-effect insights. It translates complex results into clear contrasts and predictions for exploratory grids and by-group analyses.