Back to search

University of Washington

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

Abstract

dc:description.abstract

Reporting effect sizes is an important staple of modern research, as they facilitate the quantification of variable relations for meta-analyses irrespective of operational scales used, enable policymakers to understand practical returns on a program investment, help alleviate reliance on null hypothesis testing, and are an essential ingredient for planning research sample sizes. In single-level logistic regression, it is well-known that traditional variance-explained metrics like those in linear regression are not possible; instead, pseudo-R^2 measures that capture how much better a model with regressors fits compared to a null model are in wide use and are available in popular statistical computing packages. While Nakagawa et al. (2017) have proposed a McKelvey-Zavoina-type metric, there have been no systematic studies of effect sizes for multilevel (random effects) logistic regression models. Further, our systematic review of educational and psychological applied literature over the past five years indicates that most research employing multilevel logistic models does not report any effect sizes – likely due to the gap in the methodological literature. The present study therefore uses simulation to compare the performance of pseudo-R^2 metrics for multilevel logistic regression models, including the McKelvey-Zavoina proposed by Nakagawa, as well as a new adjusted McKelvey-Zavoina metric, for 2-level random intercept models with varying ICCs, numbers of clusters, and cluster sizes. Those results indicate that the adjusted-McKelvey-Zavoina metric was best at reproducing the underlying data-generating R^2 value. Limitations and future research directions are discussed.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Zheng, Zixie
Advisor dc:contributor.advisor
  • Sanders, Elizabeth

Subjects

dc:subject × 8

Rights

dc:rights
Statement dc:rights
  • CC BY-NC-ND
Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1773/53532

Chain of custody

source
Harvested from
University of Washington
Base URL
digital.lib.washington.edu/server/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Zheng, Zixie. Comparing Pseudo-R-squared Metrics for Multilevel Logistic Regression Models. 2025. https://hdl.handle.net/1773/53532