基于EmpowerStats的混杂因素筛选及其校正方法  

Selection and adjustment of potential confounders based on changes of effect size using EmpowerStats

作　　者：施红英[1] 陈常中[2] 毛广运[1] 黄陈平[1] 杨新军[1] 

机构地区：[1]温州医科大学公共卫生与管理学院预防医学系,浙江温州325035 [2]美国哈佛大学医学院Dana.Farber癌症研究所,马萨诸塞州02115

出　　处：《温州医科大学学报》2017年第5期361-365,共5页Journal of Wenzhou Medical University

基　　金：国家自然科学基金青年基金资助项目(81502893);浙江省公益性技术应用研究计划项目(2014C33160);浙江省教育厅科研基金资助项目(Y201327770)

摘　　要：目的:介绍和演示一种新的混杂因素筛选和校正方法。方法:从原理简介、实例讲解、软件操作多角度全面介绍如何根据粗效应值和调整效应值的变化实现混杂因素的筛选以及独立效应评价。结果:Empower Stats统计软件能够按照一定的标准,科学、简便地实现混杂因素的识别、筛选及其控制,得到对效应值的最优估计,优于传统的逐步回归法。结论:基于效应估计值的改变进行混杂因素的识别和筛选,可以更合理地获得研究因素的效应估计值。Objective： To introduce a new method for selecting and adjusting confounding factors. Meth-ods： The disadvantage of traditional method for selecting confounders including methods based on P value or stepwise regression was analyzed was analyzed, and a new method based on the change of effect size was proposed to select the potential confounders which need to be controlled. And the study also demonstrated the application of EmpowerStats software using the new method. Results： EmpowerStats statistical software could automatically choose right regression methods and select the appropriate confounding factors based on the change of effect size conveniently. Conclusion： Selecting confounding factors based on the change of ef-fect size is a better choice, and can give a more accurate independent effect, and has been widely used and ac-cepted worldwide.

关 键 词：混杂因素 偏倚 协变量 统计学 

分 类 号：R195.1[医药卫生—卫生统计学]