MODELS FOR ANALYZING OVER-DISPERSED HURDLE NEGATIVE BINOMIAL REGRESSION MODEL: AN APPLICATION TO MANUFACTURED CIGARETTE USE

Abstract

Our main aim is to identify the factors that influence the use of manufactured cigarettes among tobacco users especially those whose age is above fifteen. Among the tobacco users, a large portion of adult does not take manufactured cigarettes but take other tobacco. As a result, we need to construct a model that can handle the existence of excess zero counts and the over-dispersed phenomenon. Motivated by these facts, in this paper, we propose to apply the Hurdle Negative Binomial (HNB) regression model to discover the relationships between uses of manufactured cigarettes and social factors. The data were found to have excess zeros (35%); moreover, the variance is 47.122, which is much higher than its mean 5.933. With excess zeros and high variability of non-zero outcomes, the HNB model was found to be better fitted.