Sankhya: The Indian Journal of Statistics
On Generalized Quasilikelihood Inference in Longitudinal Mixed Model for Count Data
Brajendra C. Sutradhar and Wasimul Bari, Memorial University of Newfoundland, St. John's, Canada
SUMMARY. In longitudinal studies for count data, a small number of repeated count responses along with a set of covariates are collected from a large number of independent individuals. In this set up, it is traditionally assumed that the repeated responses are serially correlated. In this paper, we consider an extended model where it is assumed that the repeated responses are also influenced by an individual random effect causing the repeated responses to be overdispersed as well as serially correlated. Through a simulation study, we examine the performance of the generalized quasilikelihood (GQL) estimation approach in estimating the regression effects of the time dependent covariates on the responses, the overdispersion parameter, as well as the serial correlation parameter. Various misspecification effects such as effects of ignoring the serial correlations, overdispersion and non-stationarity in serial correlations, are also studied. The estimation methodology is illustrated by analysing a real data on health care utilization.
AMS (2000) subject classification. Primary 62H12; Secondary 62F12, 62P10.
Key words and phrases. Consistency, efficiency, generalized quasilikelihood estimation, Gaussian random effects, non-stationary longitudinal correlation structure, repeated count response.