Strict Arcsine Regression for Longitudinal Count Data: A Marginal Modelling Approach

The strict arcsine model of the natural exponential family with cubic variance function is a feasible alternative for statistical analysis of overdispersed count data. Following from this development, a strict arcsine regression model is proposed for applications in explanatory and predictive modelling. In this paper, we developed a strict arcsine regression model for correlated count data via the generalized estimating equation approach. An application of the developed model is illustrated on a longitudinal data set on epilepsy treatment.