Sample size for clustered count data based on discrete Weibull regression model

Abstract

In this study we consider the problem of sample size determination for clustered count data based on clustered discrete Weibull regression model. Many papers that deal with sample size calculation for clustered count data have a strong assumption about the dispersion type of the data. Poisson, negative binomial are widely used models however their assumptions are hard to be met in real life. Here we propose to use discrete Weibull regression model which can handle both under and over dispersed type of data for clustered count data when calculating the sample size. We incorporate a random intercept to consider the correlation between the subjects in each cluster. We used the h-likelihood method and Monte Carlo simulation to estimate the required sample size for clustered count data. Extensive simulation study is also performed to examine the effect of the skewness of the data, variance of the random effect and the number of the clusters on calculating the required sample calculation.