心理与教育元分析中相关系数的范围限制矫正

Correction for Range Restriction for Correlation in Psychology and Education Mate-analysis

介绍了如何利用矫正公式在元分析中对相关系数的范围限制进行矫正.CaseⅠ和CaseⅡ针对直接范围限制,而CaseⅢ、CaseⅣ和CaseⅤ针对间接范围限制.由于条件较难满足CaseⅠ和CaseⅢ很少被使用,CaseⅡ则是早期运用最多的矫正公式.因现实研究多数为间接范围限制,针对直接范围限制的CaseⅡ会低估相关系数真实值,这种情况下CaseⅣ和CaseⅤ比CaseⅡ准确.CaseⅣ和Case V也有弱点,CaseⅣ的核心假设,即第三方变量S对效标变量P的影响全部由预测变量T中介,通常无法满足,而当预测变量和效标变量的相关接近0时,就难以选择CaseV的矫正公式.

The purpers of this paper is to intruduce correction equation for range restriction.Case I and Case II are correction for direct range restriction,while CaseⅢ,CaseⅣ and CaseV for indirect range restriction.Case I and CaseⅢ can almost never be used because the conditions or informations which they require would rarely be met or available.Instead,CaseⅡ which assums direct selection on X has been almost universally used.However,it has been shown that use of the CaseⅡ correction leads to underestimates the true validity.Studys indicate that the CaseⅣ and CaseⅤ procedure for correction for indirect range restriction is more accurate than Case Ⅱ.The key assumption of CaseⅣ,that the effect of third variable S on which seletion actually occurred on the ture sore P is fully mediated by the true score T,cannot be verified.And it will be to difficult to determine which eqution of CaselV should be used when the population correlation between T and P is close to zero.