easystats/performance

What it solves

It provides a consistent and unified way to evaluate the quality of statistical regression models. Instead of using fragmented tools across different packages, it offers a single interface to compute fit indices and perform diagnostic checks to ensure a model is a good fit for the data.

How it works

The package provides a suite of functions that analyze model objects (including linear, logistic, mixed-effects, and Bayesian models). It calculates performance metrics like R-squared and RMSE, runs diagnostic tests for issues like overdispersion or singularity, and allows for the direct comparison of multiple models using composite performance scores.

Who it’s for

Researchers and data analysts using R who need to validate, test, and compare the goodness-of-fit for various types of statistical regression models.

Highlights

• Unified Fit Indices: Generic functions like r2() and icc() that work across multiple model types.
• Model Diagnostics: Specialized checks for overdispersion, zero-inflation, singularity, and heteroskedasticity.
• Comprehensive Visualization: The check_model() function provides a visual overview of model assumptions.
• Model Comparison: Tools to compare several models side-by-side and rank them using a composite performance score.