Comparing Generalized Linear Mixed Models and Generalized Estimating Equations with Longitudinal Count Data: A Demonstration

• In Event: Research in Progress Roundtable Session 3
  In Invited Roundtable: Education Research Methods: Narrative Data Collection, Experimental Design, Sequential Explanatory Mixed Methods, and Longitudinal Count Data Analysis (Table 17)

Abstract

Longitudinal count data posits unique challenges for applied researchers to choose appropriate models and provide correct interpretations. This study compares Generalized Linear Mixed Models (GLMM) and Generalized Estimating Equations (GEE) in analyzing longitudinal count data, which provides conditional (subject-specific) and marginal (population-average) estimates, respectively. Using data from the National Longitudinal Survey of Youth (NLSY97), we fitted GLMM and GEE models to examine the trend and sex differences regarding the number of drinks consumed. Findings from both models underscored similar and significant trend change in drinking behavior, and confirm that gender significantly influences drinking patterns. The similarities and differences in the results between GLMMs and GEEs were graphically illustrated and discussed.

Authors

• Jichuan Wu, University Of Minnesota

• Haoran Li, University of Minnesota