Designs for fitting Poisson regression model

Abstract

Experiments related to herbicides or insecticides usually have the objective to find the effective concentration of the chemicals to control weeds or insects and to understand the relationship between the response and explanatory variables. The response is the number or proportion of organisms died and thus, is count data. The present study deals with the problem of developing experimental designs under Poisson regression model, which is a nonlinear model with count data as response. The focus here is to determine the unknown parameters of the model efficiently. The statistical designs generated are saturated and their performance is found better than traditionally used equally spaced designs. A simulation study is presented to demonstrate the application of the generated designs in actual experiment.