数值变量正态性检验常用方法的对比  

The Comparison among the Common Normality Tests for Numerical Variables

作　　者：马兴华[1] 张晋昕[1] 

机构地区：[1]中山大学公共卫生学院,广州510080

出　　处：《循证医学》2014年第2期123-128,共6页The Journal of Evidence-Based Medicine

基　　金：中山大学本科教学改革研究课题资助项目[中大教务(2012)173号];广东省高等教育教学改革重点资助项目[粤教高函(2013)113号]

摘　　要：正态分布是一种重要的连续型概率分布。统计分析中,研究者们时常跳过了对所研究变量的正态性检验,直接认为资料满足正态假定。如果恰逢涉及的数值变量并非来自正态总体,就会得到错误的分析结果。常用的正态性检验方法主要有两大类:一是主观判断的图示方法,二是客观量化的统计指标计算辅以检验。本文介绍常见的几种正态性检验方法,为数值变量统计学分析方法的正确运用提供参考。Normal distribution is one of the important distributions for continuous variables. In many statistical analyses, researchers often assumed directly the data were normally distributed without test before hand. If the numerical variables are not normally distributed, it may lead to wrong results. Normality tests which are commonly used can be classified into two categories. One category includes subjective graphical methods , and the other category contains computational methods including hypothesis tests which are objective and quantitative. In this paper , several normality tests popularly used will be introduced according to the above two categories to provide references for proper applications of statistical analyses of numerical variables.

关 键 词：正态分布 正态检验 适用条件 

分 类 号：R195.1[医药卫生—卫生统计学]