rr2: R2s for Regression Models

title: 'rr2: An R package to calculate $R^2$s for regression models' tags: - R - $R^2$ - GLMM - phylogenetic regression - models with correlated data authors: - name: Anthony R. Ives orcid: 0000-0001-9375-9523 affiliation: 1 - name: Daijiang Li orcid: 0000-0002-0925-3421 affiliation: 2 affiliations: - name: Department of Integrative Biology, UW-Madison, Madison, WI 53706 index: 1 - name: Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, FL 32611 index: 2 date: 6 September 2018 bibliography: paper.bib

Summary

Reporting the variance explained by a model (an $R^2$) is common for many simple statistical tests. However, conceptual challenges exist in defining $R^2$ for models that include correlated data. @ives2018r2s proposed three $R^2$s ($R^2_{lik}$, $R^2_{resid}$, and $R^2_{pred}$) for a variety of regression models that include correlation among data such as linear mixed models (LMMs), generalized linear mixed models (GLMMs), and phylogenetic regressions [PGLMMs, @ives2011generalized; @ives2014phylogenetic]. These three $R^2$s can also be used as partial $R^2$s to compare the contributions of predictor variables (fixed effects) and/or correlation structures (random effects) to the fit of the models.

The rr2 package provides R functions to implement the $R^2$s proposed by @ives2018r2s. The main function, R2(), calculates all three $R^2$s by default, with arguments available to select which $R^2$(s) to calculate by users. Alternatively, individual $R^2$s can be calculated with corresponding functions (R2_lik(), R2_resid(), and R2_pred()). Supported models include linear models (lm), generalized linear models (glm), linear mixed models (lmerMod), generalized linear mixed models (glmerMod), phylogenetic generalized least squares models (phylolm), phylogenetic logistic regression (phyloglm), and phylogenetic generalized linear mixed models (binaryPGLMM and communityPGLMM).