A1174
Title: A flexible approach to geographically weighted modeling for count data
Authors: Vivian Yi-Ju Chen - National Chengchi University (Taiwan) [presenting]
Abstract: Discrete count data frequently arise in spatial applications and often exhibit complex features, including overdispersion, underdispersion, and zero inflation. In response, various geographically weighted count models (GW count models) have been proposed to simultaneously address these data characteristics and capture spatially varying relationships in spatial count data. However, such techniques typically rely on specific distributional assumptions and model formulations, which can lead to inconsistent interpretations or substantial computational burdens when fitting and comparing multiple competing models. To address these challenges, a unified and flexible approach is proposed for spatial count data modeling by extending the Poisson-Tweedie family within the geographically weighted regression (GWR) framework. This approach accommodates a wide range of dispersion patterns and tail behaviors while allowing for locally varying relationships, thus avoiding the need for explicit model selection among competing GW count models. The model specification is formally presented, and relevant modeling issues are discussed, followed by an evaluation of its performance through simulations. Finally, the proposed method is applied to dengue data in Taiwan as an empirical illustration.
