|
Abstract:
|
Excess-zero models are used for count data that have larger number of zeros than expected by traditional count distributions. One of the common distributions that is used to model such data is the Poisson hurdle distribution. Poisson hurdle distribution was used to model cross sectional as well as longitudinal outcomes. When modeling longitudinal data, a common assumption is that time intervals for measuring the outcomes are the same across all subjects. However, it is occasionally seen in practice that the measurement periods are irregularly taken for different reasons. In this situation, applying the traditional statistical methods may yield to biased results. Recently, researchers started to model these type of longitudinal outcomes using what is known as joint models with informative time. In this study, a joint hurdle model for longitudinal data and informative time is constructed. Parameter estimation using the method of maximum likelihood is shown and the asymptotic properties are studied through simulation. Moreover, the likelihood ratio test statistic and some model selection criteria are computed. Lastly, an application is presented and data were analyzed for demonstratio
|
