講  題:Exploring spatial nonstationarity for continuous nonnegative response data

主講人:陳怡如 副教授(國立政治大學統計學系)

時  間:20231109日(星期四)下午0210 - 0400

地  點:B302A(淡水校園商管大樓)

茶  會:20231109日(星期四)下午0130 (商管大樓 B1102)

 

摘 要

 

    Geographically weighted regression (GWR) has been a popular tool applied in many disciplines to explore spatial nonstationarity (or heterogeneity) with respect to data relationships for georeferenced data. However, GWR is typically limited to analyzing continuous dependent variables assumed to follow a symmetric normal distribution. In many fields, nonnegative continuous data are frequently observed and may come with substantial amounts of zeros followed by a right-skewed distribution of positive values. When dealing with such type of outcomes, GWR may not provide adequate insights regarding spatially varying regression relationships. This study intends to extend GWR based on compound Poisson distribution, thus allowing for not only the exploration of relationship heterogeneity but also the accommodation of spatial nonnegative continuous response variables. We first present the model specification of the proposed method and then discuss the associated modeling issues, such as bandwidth and tests for spatial nonstationarity. We evaluate the performance of this new technique through simulations. Finally, we conclude the study with an empirical illustration based on a dataset of dengue fever in Tainan, highlighting the applicability and utility of the proposed approach.


建議最佳瀏覽 Microsoft IE 10 以上/Google Chrome/Mozilla Firefox 或相容W3C網頁標準之瀏覽器

更新日期 : 2024/06/11