分位数回归在比例数据分析中的应用

Application of the Quantile Regression in the Proportional Data Analysis

基于分位数方法研究了比例数据的回归建模、估计、推断及其应用问题.由于比例数据具有异方差、多峰性等特点,与均值回归相比,使用分位数回归不仅能动态地描述比例响应变量与协变量之间的相关关系,而且能够揭示变量之间内在的规律和提供更加丰富的信息.数值模拟研究和家庭食物消费数据分析说明了所提方法的有用性.

Due to the features of proportional data such as heteroscedasticity and multimodality, in contrast with mean regression, quantile regression can not only dynamically describe the correlation between response variable and covariates but also unveil their innate laws and provide more information. In this study, the regression modeling of the proportional data and its estimation and inference method were probed, and the proposed method was employed to real data. Simulation studies and household food expenditure data analysis were used to illustrate its usefulness.