Analysis of Overdispersed Count Data: an Application on Acar (Acarina) Counts

Abstract

Overdispersed count data sets are frequently encountered in plant protection area as in various fields of study. This kind of data set sometimes has more zero values than expected in accordance with Poisson distribution. This study aims to determine the regression method with best models the numbers of protogyne, deutogyne, nymphopupa and eggs of Aculus schlechtendali (Apple Rust Mite; Acarina: Eriophyidae), known to cause serious economic loss on Fuji, Gala, Red Chief, and the Granny Smith apple produce of Van province. To serve this purpose, Poisson, Negative Binomial, Zero Inflated Poisson, and Zero Inflated Negative Binomial regression models were applied for statistical analysis. Akaiki Information Criteria and Bayesian Information Criteria were used in order to determine the best model. The best model was determined as Zero Inflated Negative Binomial regression for the numbers of protogyne, as well as Negative Binomial regression for the numbers of deutogyne, nymphopupa and eggs.