HKIER

摘要

本文討論應用結構方程模式一些較重要問題和常見謬誤，包括：（一）數據與模式擬合的意義、（二）等同模式問題、（三）擬合優指數、（四）模式修訂及交互效度、（五）相關與因果關係、（六）潛伏變項之名實相符問題、和（七）用SEM分析相關矩陣問題。本文綜合並重申學者指出運用SEM常見之謬誤，亦嘗試提出解決辦法，供研究者參考。

Abstract

We discuss a number of common issues and misconceptions in using structural equation modeling (SEM). They include: (1) the concept of goodness of fit, (2) the issue of model equivalency, (3) the choice of fit indexes, (4) model modification and cross-validation, (5) correlation and causality, (6) labelling of latent constructs, and (7) using correlation matrices in SEM. We summarize and reiterate common mis- conceptions and concerns in SEM analyses and attempt to propose strategies for solving the problems.