Application of INAR model on the pest population dynamics in Agriculture

Abstract

In real world phenomenon, some situations often occur where observations are not continuous. Such situations in time series may be number of road accidents, number of patients hospitalized, number of phone calls, the number of products sold, etc. In agriculture, it may be number of pests emerged in a crop season, number of crops infected by a certain type of disease. In these situations, usual autoregressive (AR) model where errors follow Gaussian distribution is not applicable. Hence there is a need to develop models for count observations. The frequently used model for count observation is Poisson regression. In this paper, Integer valued autoregressive (INAR) model is applied for prediction of count observations. Here two real data sets are applied where prediction has been done on pest populations. Finally INAR model is compared with Poisson regression and it is seen that INAR model gives better forecast than the Poisson regression model.