数据缺失机制识别及处理的标准化流程及集成系统  

A standardized process and correspondingly integrated system for mechanism recognition and imputation of missing data

作　　者：岳廷妍 张昱勤 李晓松[1] 马越[1] 张韬[1] YUE Ting-yan;ZHANG Yu-qin;LI Xiao-song;MA Yue;ZHANG Tao(West China School of Public Health and West China Fourth Hospital,Sichuan University,Chengdu,Sichuan 610041,China)

机构地区：[1]四川大学华西公共卫生学院/四川大学华西第四医院

出　　处：《现代预防医学》2019年第21期3928-3932,3936,共6页Modern Preventive Medicine

基　　金：国家自然科学基金青年科学基金(No.81602935);四川大学青年教师科研启动基金(2016SCU11006)

摘　　要：目的提出数据缺失机制识别及处理的标准化操作流程,并开发相应集成系统,为非统计专业背景的医学工作者处理缺失数据提供恰当、专业且简便的实现工具。方法系统集成了完成者数据集法、K最近邻分类算法和链式方程多元插值法等缺失数据处理方法,并将其归纳到缺失机制识别及处理的统一框架下,为缺失数据处理提供了从缺失统计,缺失机制识别到缺失处理的标准化流程。结果将归纳的标准化流程分步骤开发为缺失统计、缺失识别、缺失处理等功能模块并进行了集成化,构建了缺失机制识别及处理集成系统。结论标准化操作流程及集成系统实现了缺失机制识别加缺失数据处理全过程,操作方式简单便捷,结果展示直观易懂,为缺失数据的处理提供了更为简便可行的选择,便于医学工作者实际应用。Objective To help medical workers without a statistics foundation impute missing data properly through introducing a standardized process for recognizing the missingness mechanism,dealing with incomplete data and developing a correspondingly professional and user-friendly integration system.Methods Several imputation techniques including complete case analysis,k-nearest neighbour algorithm and multivariate imputation by chained equations algorithm were integrated into a unified framework of the system named“mechanism recognition and imputation for missing data”.The system provided standardized processes that included missingness summary,missingness mechanism recognition and missing data imputation.Results The standardized processes were developed into corresponding function modules to build the integrated mechanism recognition and imputation system for missing data.Conclusion This easy and convenient integrated system has realized the whole process of recognizing missingness mechanism and imputing missing data and can present the results straightforward and understandably,which has made the missing data handling briefer and more feasible for medical workers in analysis practice.

关 键 词：缺失数据 缺失机制识别 缺失值填补 集成系统 

分 类 号：R194[医药卫生—卫生事业管理]