A parameter estimation method for structural equation model based on generalized maximum entropy

Abstract

In order to resolve the problem of parameter estimation in structural equation model, we introduce the entropy theory into parameter estimation. According to the principles of Generalized Maximum Entropy (GME), we study the results and performances of GME method and Partial Least Squares (PLS) method in solving parameters estimation of structural equation model in small-sized sample data. By comparison of two methods, we could conclude that GME method outperforms PLS method when dealing with small-sized sample data in stability and results. Finally, on the basis of this result, we provide some suggestions on solving parameter estimation of structural equation model in small-sized sample data.