| Summary: | 碩士 === 銘傳大學 === 應用統計資訊學系碩士班 === 99 === Many researchers from different fields are actively involved in the traditional Structural Equation Modeling (SEM) (classified as 1st generation, also called as the standard SEM), and tried to understand, learn and apply it. However, some scholars have already introduced the second generation of SEM (Lee, 2007). The standard SEM, in particular the LISREL model (Jöreskog and Sörbom, 1996), is composed of two components. The first component is a confirmatory factor analysis model (CFA) which consists of the latent variables to all their relating manifest variables and takes the measurement errors into account. This component can be regarded as a regression model which regresses the manifest variables with a small number of latent variables. The second component is a regression type structural equation which regresses the endogenous latent variables with the linear endogenous and exogenous latent variables.
In recent years, the growth of SEM has been very rapid. New models and statistical methods have been developed for better analyses of more complex data structures in practical research. Therefore, there is a need for the 2nd generation of SEM which involves a much wider class of SEM that include the standard SEM and their useful generations. The purpose of the study is to introduce the Bayesian SEM, and the research data is WHOQOL-BREF. After empirical testified, the analysis results from BSEM are really better than the traditional SEM, especially for factor loadings.
|
|---|
