处理缺失数据的分数插补法研究  

Study on Fractional Imputation in Dealing with Missing Data

作　　者：杨帆[1] 庞新生[1] 

机构地区：[1]北京林业大学经济管理学院,北京100083

出　　处：《统计与决策》2017年第14期15-18,共4页Statistics & Decision

基　　金：国家社会科学基金资助项目(13BTJ021)

摘　　要：随着研究中对数据质量要求的提高,缺失数据相关问题也越来越受到重视。文章主要论述了处理缺失数据的方法之一——分数插补法的理论基础,并在此基础上研究了分数热卡插补法及其方差估计,同时使用模拟数据,对分数热卡插补法的实现过程做了模拟研究。通过对比实验,可以得到分数热卡插补法能够在保证原有数据分布的基础上,减少因插补造成的偏差,提供更加准确的插补结果。With the increasing requirement of data quality in research, missing data and problems deriving from it have be- come an important concern nowadays. This paper mainly discusses fractional imputation, one of the methods dealing with missing data, on the basis of which the paper studies fractional hot deck imputation method and its variance estimation. The paper also em- ploys simulated data to study the procedure of fractional hot deck imputation. Through a comparative experiment the paper comes to a conclusion that with the original distribution of data, fractional hot deck imputation method reduces the deviation caused by interpolation and provides more accurate interpolation results.

关 键 词：缺失数据 分数插补法 分数热卡插补法 刀切法 

分 类 号：O21[理学—概率论与数理统计]