Comparing Pseudo-R-squared Metrics for Multilevel Logistic Regression Models

• In Event: AERA Roundtable Session Wednesday 3:45 pm - Gold 3
  In Roundtable Session: Recent Developments in Multilevel Modeling: Model Estimation and Evaluation

Abstract

Reporting effect sizes is an important staple of modern research. In single-level logistic regression, pseudo-R^2 measures that capture how much better a model with regressors fits compared to a null model are widely available, yet little guidance exists for multilevel logistic regression models. While Nakagawa et al. (2017) proposed one McKelvey-Zavoina (MZ) type of metric for multilevel logistic regression models, there have been no systematic studies of this metric compared to others. The present study takes up this issue by 1) exploring current effect size reporting practices for multilevel logistic regression modeling, and 2) simulating data to compare the performance of different pseudo-R^2 approaches for 2-level random intercept logistic regression models, including the MZ metric.

Authors

• Zixie Zheng, University of Washington

• Elizabeth A. Sanders, University of Washington