Joint Modeling of Mixed Responses with Bayesian Modeling and Neural Networks: Performance Comparison with Application to Poultry Data

Abstract

Joint modeling of mixed responses has become a popular research area due to its applicability in many disciplines. The interest of this study is joint modeling of survival and count data. Survival data is continuous in nature with censoring information combined to it, while count is a discrete variable. Due to this fact, joint modeling of these two variables will be a challenging task, but it will provide interesting and improved results than modeling these two variables separately. In this study, the concept of joint modeling of survival and count data has been carried out using two approaches: Bayesian modeling and Neural Networks, in order to compare their performances. The results of an application to the poultry data revealed that the Neural Network has a better fit in general.