Statistical modeling for zero-inflated data using GAMs

Abstract

Statistical modeling for zero-inflated data using GAMs Count data are common in biological studies, such as in entomology, where numbers of individuals or proportions are observed. Generally, these experiments or sampling processes have overdispersion and excess of zeros. The use of zero-inflated models supports the study process for data sets with these behaviors. The first paper presented here deals with an experiment conducted with a cotton pest. The main objective is to compare the models for zero-inflated overdispersed count data. Generalized linear models and generalized additive models are compared in terms of fit according to the inclusion of smoothing functions. After the selection of the model, covariates of natural enemies were included. Models that used smoothing functions allowed a better evaluation of the biological interactions over time between the pest and its natural enemies. The second work deals with a simulation study that aims to compare the efficiency of the inclusion of smoothing functions for simulated data sets of zero-inflated longitudinal data. The inclusion of smoothing functions did not show improvement in the fit of the models for the created simulation scenarios.