Visualizing Count Data Regressions Using Rootograms

We show how the rootogram - a graphical tool associated with the work of J. W. Tukey and originally used for assessing goodness of fit of univariate distributions - can help to diagnose and treat issues such as overdispersion and/or excess zeros in regression models for count data. Two empirical illustrations, from ethology and from public health, are included. The former employs a negative binomial hurdle regression, the latter a two-component finite mixture of negative binomial models for which weighted versions of rootograms are utilized.