在心理和行为等研究领域，使用结构方程模型(SEM)分析潜变量关系已很常见。在这些领域，多维构念(包含了多个不同但相关的维度)的测验题目往往存在跨因子负荷，因而不满足独立分群模型假设，这时使用探索性结构方程模型(ESEM)进行统计分析被认为是较好的做法。本项目要研究的是多维测量数据分析的一个前沿课题――ESEM与传统模型比较，有什么优缺点。研究重点是，分别对回归分析、中介效应分析和调节效应分析，将ESEM与SEM、MPM(显变量路径模型)做比较。目标是了解什么情况下ESEM比SEM和MPM都好，有必要使用ESEM，什么情况下使用SEM甚至简单的MPM就可以了，为应用工作者提供多维潜变量建模方法指引。对于中介分析和调节分析，还需要研究现有估计和检验方法在ESEM框架下的适用性，给出ESEM建模步骤、估计和检验策略以及相应的Mplus程序。也将用本项目新创的方法，研究两个涉及多维构念的案例。

Structural Equation Models (SEMs) have been frequently employed to study the relationship among latent variables in psychology, as well as in many other behavioral and social science disciplines, in which multidimensional constructs are often involved. A construct is multidimensional when it includes several distinct but related dimensions treated as a single theoretical concept. A measurement model with at least one multidimensional construct often violates the assumption of the independent clusters model in which each item has only one loading on the main factor because there are intercross-loadings on the other factors. In such a case, an Exploratory Structural Equation Model (ESEM) is considered more appropriate than a traditional SEM for statistical modeling. In this project, we will research statistical modeling methods under the framework of ESEM with their merits and demerits taken into account. The study will focus on path model comparison among ESEM, SEM and MPM (Manifest Path Model) in order to find out under which circumstances ESEM is better than SEM and MPM, and SEM or even simple MPM behaves well enough. The comparison will be made for regression analysis, mediation analysis and moderation analysis respectively. For the latter two relatively complicated analyses, we also need to investigate the appropriateness of the existing modeling methods within the framework of ESEM, in order to propose and recommend modeling procedures, strategies for model estimation and test, and Mplus programs. Finally, we will apply some new methods/models obtained in this project to two case studies in which multidimensional constructs are involved.