ABSTRACT
This chapter describes the applications of generalized linear mixed model (GLMM) to mixture models with random effects. These mixture models are useful to analyze multilevel data or recurrent event data, where the response outcome can be measured by a count variable or a failure-time variable with censored observations. The GLMM approach provides an advanced modelling technique to handle response variables that are dependent with complex correlation structures, as arisen from the aforementioned study designs. Assuming normally distributed random components for the random effects, the GLMM approach commences with developing the Best linear unbiased predictor estimators in the initial step and proceeds to obtain Residual maximum likelihood estimators of regression and variance component parameters. Poisson mixture regression models can be adopted to analyze multilevel count response variables, where random effects are introduced to account for the inherent correlation among observations within a hierarchical unit.
