Theoretical research on structural equation model based on principle of entropy

The theory of entropy is an effective tool to solve the uncertain problems, which has a variety of applications in social and natural sciences. Based on the theory of entropy, the basic concepts and principles of entropy and information entropy, maximum entropy and probabilistic distribution are introduced comprehensively and systematically. The modeling method of structural equation model (SEM) mainly includes the linear structure relations based on the covariance structures (LISREL) and the path method based on partial least square (PLS). These two methods have different requirements for data. LISREL requires observed variables with normal distribution, while PLS has no strict restriction on data. The theory of entropy is used in the modeling research of SEM, and the entropy model with data normal distribution is established by introducing the criterion and constraint condition of maximum entropy, and thus the normal distribution is determined. According to this the modeling method of SEM is determined. Therefore, the modeling idea based on the theory of entropy is a useful exploration for the modeling theory of SEM.