구조방정식 모형에서 χ2 검증의 문제점 및 대안탐색

초록

The purpose of this study is to address the problem of chi-square test statistic in terms of model selection and to present alternative ways of model fit in structural equation modeling. Chi-square test statistic is not a very good fit index in practice under many situations because it is affected by the following factors: 1) sample size, 2) model size, 3) the distribution of variables, 4) omission of important variables. Therefore, the chi-square test statistic should not be the sole basis for determining model fit. As a result of this situation, many alternative measures of model fit have been proposed. I consider one such measure called Root Mean Square Error of Approximation, or RMSEA here. A major advantage of RMSEA over many other fit measures is the availability of confidence intervals for RMSEA and the testing the null hypothesis of close fit.