The past years have experienced growth in the methodological development that
intend to explore spatial nonsationarity for spatially count data based on the technique of geographically weighted regression. Several geographically weighted count models have been introduced in literature to deal with the challenges of analyzing the count data without/with overdispersion and/or excessive zeros. However, researchers have lagged to provide a comparative assessment across all the proposed methods. In this study, we argue that spatial analysts should pay sufficient attention to analytical model comparisons since different geographically weighted count models may generate competing accounts of the same data set. Here we first review the existing techniques and introduce geographically weighted zero-inflated negative binomial model as a methodological complement. Several qualitative measures and graphical tools are then suggested to compare among various GW count models. We also illustrate their utility using an example from a study of Taiwan dengue data. The results demonstrate the importance of model comparisons in investigating spatial nonstationarity for spatial count data analyses.
