结构方程模型简化方法及在人机行为研究中的应用

为解决研究中结构方程模型过度复杂的问题,提出一种简化方法.分析了多维测量模型直接引入整体模型的局限性,对高阶验证性因子分析方法和项目组合法在模型简化中的适用性进行了分析,提出了首先检验简化必要性再使用高阶验证性因子分析方法和项目组合法对整体模型进行简化的两阶段方法.以电子学习系统持续使用行为的研究数据为例对该简化方法的有效性进行检验,简化前后拟合数据的对比分析表明该方法可以显著简化整体模型并获得理想的拟合优度.

Simplification and Application of Structural Equation Model in Human Computers Behavioral Research

To reduce the over complexity of a structure equation model in researches,we put forward a paradigm for the simplification of SEM.We first analyzed limitations of introducing a multidimensional measurement model directly into an integrated one,and then elaborated the applicability of employing approaches as higher-order confirmatory factor analysis and item parceling in the model simplification.The paradigm examined the simplification necessity firstly and then combined the two approaches in the simplification.Taking the research of users' retention on e-leaming systems as an example,we verified the effectiveness of the paradigm.The comparisons on data between the original model and the simplified one reveal that the paradigm is effective in reducing model complexity with idea goodness of fit.