Análise de diagnóstico em modelos de regressão Bell

Abstract

The class of Bell regression models is an interesting alternative to the usual class of Poisson regression models for response variables in count form, mainly in the presence of overdispersion. In this work, diagnostic techniques such as global influence and local influence will be developed for the Bell regression model. In particular, analytical expressions for normal curvature to study local influence will be obtained under different perturbation schemes. Finally, the diagnostic techniques developed will be applied to a real data set.