Confronting Quasi-Separation in Logistic Mixed Effects for Linguistic Data: A Bayesian Approach

TL;DR

This paper addresses the problem of quasi-separation in logistic mixed effects models for linguistic data, demonstrating how Bayesian methods can effectively overcome convergence issues that frequentist approaches struggle with.

Contribution

It introduces a Bayesian approach to handle quasi-separation in logistic mixed effects models, improving convergence and estimation accuracy in linguistic research.

Findings

Bayesian models solve convergence issues caused by quasi-separation.

Frequentist methods often fail in the presence of quasi-separation.

Bayesian approach provides more reliable estimates for linguistic binary data.

Abstract

Mixed effects regression models are widely used by language researchers. However, these regressions are implemented with an algorithm which may not converge on a solution. While convergence issues in linear mixed effects models can often be addressed with careful experiment design and model building, logistic mixed effects models introduce the possibility of separation or quasi-separation, which can cause problems for model estimation that result in convergence errors or in unreasonable model estimates. These problems cannot be solved by experiment or model design. In this paper, we discuss (quasi-)separation with the language researcher in mind, explaining what it is, how it causes problems for model estimation, and why it can be expected in linguistic datasets. Using real linguistic datasets, we then show how Bayesian models can be used to overcome convergence issues introduced by quasi-separation, whereas frequentist approaches fail. On the basis of these demonstrations, we advocate for the adoption of Bayesian models as a practical solution to dealing with convergence issues when modeling binary linguistic data.