A0709
Title: General Bayesian quantile regression of count via generative modeling
Authors: Yuta Yamauchi - Nagoya University (Japan) [presenting]
Genya Kobayashi - Meiji University (Japan)
Shonosuke Sugasawa - Keio University (Japan)
Abstract: A novel Bayesian framework is presented for estimating quantile regression functions of the discrete response based on the inference of the conditional cumulative distribution of the discrete response. The approach involves the following steps: first, the joint distribution of the discrete response and the covariates is estimated using nonparametric mixture methods. Next, the posterior samples of the conditional quantiles are obtained based on the induced conditional cumulative distribution of the response given the covariates from the estimated joint distribution. Finally, the relationship is represented between the conditional quantile and the covariates using an additive model. The posterior samples of the quantile regression function based on this additive model can be obtained by minimizing the difference between the sampled conditional quantiles and the regression function using a loss function. Through simulation studies, the high flexibility of the method is demonstrated in capturing the relationship between the conditional quantiles and the covariates.
