Modeling Relative Risk Assesment for Infected Plants by Eriophyoid Mites (Acari, Prostigmata) Using Poisson Log Linear Regression Model

Abstract

In Poisson regression, the dependent variable of the mite population relative risk assessment can be estimated as based on countable data. Due to disappearing data and numerous uncountable values of the mite population it became difficult to evaluate the risk factor by linear regression methods. The assessment of variable features of mites depending on conditions is very suitable for Poisson regression modeling system. In the this study, the occurrence of rare events such as the occurrence ratio of infected plants was defined by eriophyid mites on two wheat varieties and four different localities. The study was constructed by two way possibility table depending on plant varieties (Triticum aestivum L. and Secale cereale L. (Poaceae) with four locations (Muradiye, Ahlat, Erciş, Doğu Beyazıt and Iğdır). The reference parameters were Triticum aestivum for varieties, and Muradiye for location, respectively. The risk assessment of infected plants for Secale cereale is 1.245 times higher as compared to Triticum aestivum and this difference was found statistically significant (p<0.05). The risk of infected plants for Iğdır location is 1.101 times higher as compared to Muradiye location (p>0.05). In the Poisson log-linear regression, the dependent variable is a risk ratio or a relative risk can be estimated as well as countable data. Thus, Poisson log-linear regression model is a very effective method for analysis of two-way contingency table. Two-way contingency table is created considering the eriophyid mite infection ratio depending on location and varieties, respectively.