Correlation structure and variable selection in generalized estimating equations via composite likelihood information criteria

摘要

The method of generalized estimating equations (GEE) is popular in thebiostatistics literature for analyzing longitudinal binary and count data. Itassumes a generalized linear model (GLM) for the outcome variable, and aworking correlation among repeated measurements. In this paper, we introduce aviable competitor: the weighted scores method for GLM margins. We weight theunivariate score equations using a working discretized multivariate normalmodel that is a proper multivariate model. Since the weighted scores method isa parametric method based on likelihood, we propose composite likelihoodinformation criteria as an intermediate step for model selection. The samecriteria can be used for both correlation structure and variable selection.Simulations studies and the application example show that our methodoutperforms other existing model selection methods in GEE. From the example, itcan be seen that our methods allow for correct analysis, and may change theinferential results.