Robust diagnostics for the negative binomial regression model

Abstract

Modeling count variables is a common task in econometrics, social and medical sciences. The negative binomial (NB) regression model is one of the popular approaches to the fitting of overdispersed count data. However, outliers may have some effects on the maximum likelihood estimates of the regression coefficients for NB regression model. We apply the maximum trimming likelihood estimation to deal with outlier problem for the count regression model. Real data examples are used to illustrate the performance of the proposed approach.