Using Zero-Inflated Generalized Poisson Regression in Modelling of Count Data

Abstract

In this study zero-inflated generalized Poisson regression was applied to the modelling of mite numbers data based on count. The subjects of the zero-inflated generalized Poisson regression are three parameters as mean, overdispersion and zero-inflated dispersion. The overdispersion and zero-inflated dispersion levels range was obtained to be quite high. However, it was found that zero-inflated data and overdispersion had an important effect on mite counts (p < 0.01). It was obtained that 36% (130 observations) of the total numbers of mite had zero values. The effects of all independent variables were found to be statistically significant on mite counts (p < 0.05). The results showed that the differences among regions and varieties regarding the mite counts were statistically significant (p < 0.01). © 2017, Centenary University. All rights reserved.