探索性因素分析在测验编制中局限性的模拟实验

A MONTE CARLO STUDY OF THE LIMITATION OF EXPLORATORY FACTOR ANALYSIS IN TEST CONSTRUCTION

本文主要用模拟研究的方法,通过生成拟合优度验证性因素分析的数据,来考察探索性因素分析在测验编制中的局限性。结果表明探索性因素分析作为纯数据基础上的一种统计方法,在因素间相关程度较大时,得到与理论假设不一致的结论。本文还就测验中会聚效度的一些限制作了初步探讨,结合具体情况介绍了中等相关限定条件的实质。

This article, using well- fitted Monte Carlo data produced by confirmatory factor analysis, intends to inspect the limitation of exploratory factory analysis in test construction. It argues that when the factors are highly correlated, the conclusion based on EFA is not coincident with the theory postulated. In this respect, exploratory factor analysis is a completely data-oriented statistic method. The article also discusses the essence of constrictive moderate-correlation conditions of convergent validity.