구조방정식모형의 검증방법에 대한 제고찰 = A review on assessment methods of fit for structural equation model

다국어 초록 (Multilingual Abstract)

The purpose of this study is to review various assessment of fit structural equation model(SEM). SEM is considered to be the most powerful technique to root out causalities among variables, but even with the power and flexibility of the SEM, a number of limitations still remain. Especially, various assessment methods of fit have strength and weakness.
There are overall model fit, measurement model fit, and structural model fit in the goodness-of-fit criteria. First, overall model fit is a measure of the correspondence of the actual or observed input(covariance or correlation) matrix with that predicted from the proposed model. Second, measurement model fit is to assess the measurement of each construct for unidimensionality and reliability. Third, structural model fit is to assess the significance of estimated coefficients.
The goodness-of-fit measures fall into three types as follows : absolute fit measures, incremental fit measures, or parsimonious fit measures.
The absolute fir measures assess only the overall model fit with no adjustment for the degree of overfitting. Among the absolute fit measures commonly used to evaluate SEM are the Chi-square statistic, the noncentrality parameter, the goodness-of-fit statistic, the root mean square error, the root mean square error of approximation, and the expected cross-validation index.
The incremental fit measures compare the proposed model to a comparison model specified by the research through adjusted goodness-of-fit index, Tucker-Lewis index, normed fit index, and so on.
Finally, the parsimonious fit measures adjust the measures of fit to provide a comparison between models with differing numbers of estimated coefficients, the purpose being to determine the amount of fit achieved by each estimated coefficient. There are parsimonious normed fit index, parsimonious goodness-of-fit index, normed chi-square, and Akaike information criterion.
It is concluded that no single measure or set of measures have been agreed upon as the only measures needed. Therefore the researcher must choose one or more measures according to sample size, the quality and distribution of variables, the levels of acceptable fit, etc.

목차 (Table of Contents)

• 目次
• Ⅰ. 서론
• Ⅱ. 구조방정식의 개념
• 1. 측정모형
• 2. 구조모형
• 3. 구조방정식모형의 분석 절차
• Ⅲ. 모형 평가
• 1. 전반적인 모형 적합도
• 2. 측정모형 적합도
• 3. 구조모형 적합도
• Ⅳ. 모형 적합도 검증방법
• 1. 절대 적합도 측정치
• 2. 증가 적합도 측정치
• 3. 간명 적합도 측정치
• Ⅴ. 결론 및 제언
• 참고문헌
• Abstract