基于混合分位数回归的居民收入分布影响因素的统计推断  

Statistical Inference of Influencing Factors of Residents’ Income Distribution Based on Mixed Quantile Regression

作　　者：周雪娇 刘鹤飞 Zhou Xuejiao;Liu Hefei(School of Statistics,Shandong University of Finance and Economics,Jinan 250014,China;School of Mathematics and Statistics,Qujing Normal University,Qujing Yunnan 655011,China)

机构地区：[1]山东财经大学统计学院,济南250014 [2]曲靖师范学院数学与统计学院,云南曲靖655011

出　　处：《统计与决策》2022年第17期47-51,共5页Statistics & Decision

基　　金：云南省教育厅科研项目(2022J0810);山东省自然科学基金青年项目(ZR2020QG029);山东省社会科学基金青年项目(20DJJJ05)。

摘　　要：分位数回归对响应变量的分布形式没有限制,对异常值不敏感,并且可以研究不同分位点处响应变量和自变量之间的关系,因此分位数回归是研究收入分布影响因素重要的统计工具。文章针对我国居民收入分布呈现双峰的统计特征以及近年来用混合模型研究收入分布的趋势,提出混合分位数回归模型,利用混合分位数回归模型对中国家庭追踪调查(CFPS)数据进行建模,研究收入分布的影响因素。所提出的混合分位数回归模型为分布形式未知的“多峰”观测数据提供了一种新的统计建模和分析方法。Quantile regression has no restrictions on the distribution form of response variables;it is not sensitive to outliers,and can study the relationship between response variables and independent variables at different quantiles.Therefore,quantile regression is an important statistical tool to study the influencing factors of income distribution.Aiming at the double-peak statistical characteristics of residents’income distribution in China and the trend of income distribution research in recent years with a mixed model,this paper proposes a mixed quantile regression model,then uses the mixed quantile regression model to construct the China Family Panel Studies(CFPS)data,and finally makes studies on the influencing factors of income distribution.The proposed mixed quantile regression model provides a new statistical modeling and analysis method for“multi-modal”observation data with unknown distribution form.

关 键 词：混合分位数回归 贝叶斯推断 收入分布 影响因素 

分 类 号：F224.7[经济管理—国民经济]