A definite linear algorithm for structural equation model

A definite linear algorithm for structural equation model (SEM) is proposed that is developed from modular constraint (the length of vector is 1) least squares (MCLS) solution of SEM which may be used as the best iterative initial value of partial least square (PLS) algorithm. Based on MCLS, prescription regression (all regression coefficients are nonnegative and their sum is 1), and generalized linear regression model (the dependent variable is unknown), a definite linear algorithm is constructed. The data example of SEM with multilayer construction is provided to illustrate the algorithm.