A0491
Title: Geographically weighted logistic quantile regression
Authors: Vivian Yi-Ju Chen - National Chengchi University (Taiwan) [presenting]
Yu-Ting Lu - Tamkang University (Taiwan)
Abstract: In spatial analysis, geographically weighted quantile regression (GWQR) has been proposed as a method for simultaneously capturing the heterogeneous relationships through the specification of varying regression parameters and modeling the response heterogeneity with the dependent variable modeled as a series of pre-specified quantiles. However, GWQR is designed only for unbounded continuous dependent variables. When handling bounded outcome data within a defined range, GWQR may yield biased results, such as incorrect inferences and out-of-range predictions. To address this limitation, an improved model is presented that integrates the concept of logistic quantile regression. The model specification of the proposed approach is presented, and relevant modeling issues are discussed, followed by an evaluation of its performance through simulations. Furthermore, the new approach is applied to a real dataset as an empirical illustration. The analysis results demonstrate that the proposed method provides more robust and informative insights compared to other existing analytical techniques.
