浅谈线性回归分析中R-检验在应用上的错误

Error in the Applications of Attest in Linear Regression Analysis

在一元线性回归分析中,对回归效果进行检验,不仅可以用F-检验法,也可以用相关系数检验法。在多元线性回归分析中也是如此。但是,复相关系数R的临界值表与一元中相关系数r的临界值表在构成时是不一样的。然而,文献[2]却肓目地用一元相关系数的临界值表去解决多元复相关系数的临界值,产生了不应有的错误。本文就此提出看法,以便纠正这个错误。

In simple linear regression analysis as well as in multiple regression analysis, the regression effects can be tested not only by the F-test but also by the R-test, However, the structure of the critical value table of complex correlation R is different from that of simple linear correlation r. In previous research, problems of the critical value of complex correlation were resolved using the critical value table of simple linear correlation. This leads to a serious error. This paper forwards a view in order to correct the mistake.