Title: Geographically weighted quantile regression for count data
Authors: Vivian Yi-Ju Chen - Tamkang Universiy (Taiwan) [presenting]
Abstract: Past decade have witnesses an explosion both in the applications of Geographically Weighted Regression (GWR) and of Quantile Regression (QR). While these two techniques have become the commonplaces in many disciplines, they have never been integrated until an analytic framework called geographically weighted quantile regression (GWQR) has been proposed recently. The current structure of GWQR is however restricted to the analysis of continuous dependent variables. Discrete count data are observed in many fields such as health (disease counts), transportation (accidents), and finance (number of bankruptcy). When it comes to model such type of outcome, GWQR is inappropriate and provides insufficient information of the data. We aim to address the gap by extending GWQR for continuous dependent variables to the generalized GWQR framework for count variables. We first formulate the modeling specification, and then develop bootstrap methods to conducting the inference of model parameters. Finally, the proposed model is applied to a dataset of dengue fever in Taiwan as an empirical illustration.