一、问题的提出单因素完全随机设计多组计量资料均数比较,可用熟知的方差分析进行分析比较。文献中已作介绍。然而其分析过程需要计算一系列离均差平方和。在缺乏计算工具而要尽速获得多组均数比较结果时,甚感不便。本文的目的之一是介绍一种不必计算平方和的多组均数比较的简易方法——极差检验法。该法对资料概率分布的要求同传统方差分析。本文的第二个目的是介绍一种非参数检验——H检验法。这是由于统计理论要求在方差分析过程中,资料的试验误差必须满足正态性(即误差服从 First, the proposed single-factor completely random design of multiple groups of measurement data are compared, can be used to analyze and compare well-known analysis of variance. It has been introduced in the literature. However, the analysis process needs to calculate a series of squared deviations from the mean difference. In the absence of computing tools, it is very inconvenient to get multiple sets of mean comparison results as quickly as possible. One of the purposes of this paper is to introduce a simple method—a range test—that does not have to calculate the sum of squares. The law’s requirement for the probability distribution of data is the same as traditional analysis of variance. The second purpose of this paper is to introduce a non-parametric test, the H test. This is because the statistical theory requires that in the process of variance analysis, the experimental error of the data must satisfy the normality (ie, the error obeys