Let us denote the population variance/covariance matrix of variables y and x, then
(1.22)
where the diagonal elements are variances of the variables y and x, respectively; and the off-diagonal elements are covariances among y and x. In SEM it is hypothesized that the population variance/covariance matrix of y and x can be expressed as a function of the model parameters , that is:
(1.23)
where is called the model implied variance/covariance matrix.
Based on the three basic SEM equations [Equation (1.1)], we can derive that can be expressed as functions of the parameters in the eight fundamental SEM matrices. Let us start with the variance/covariance matrix of y, then the variance/covariance matrix of x and the variance/covariance matrix of y and x, and then finally assemble them together.
The variance/covariance matrix of y can be expressed as:
were is the variance/covariance matrix of the error term .
(1.25)
Assuming that is independent of , then
where is the variance/covariance matrix of the latent variable ; is the variance/covariance matrix of the residual . Substituting Equation (1.26) into Equation (1.24), we have:
(1.27)
This equation implies that variances/covariances of the observed y variables are a function of model parameters such as factor loadings , path coefficients and , the variances/covariances of the exogenous latent variables, residual variances/covariances matrix , and the error variances/covariances .
The variance/covariance matrix of x can be expressed as:
(1.28)
Assuming that is independent of , then
where is the variance/covariance matrix of the error term . Equation (1.29) implies that variances/covariances of the observed x variables are a function of model parameters, such as the loadings , the variances/covariances of the exogenous latent variables, and the error variances/covariances .
The covariance matrix among x and y can be expressed as:
(1.30)
Assuming that and are independent of each other and independent of the latent variables, then
(1.31)
Thus, the variances and covariances among the observed variables x and y can be expressed as in terms of the model parameters:
(1.32)
where the upper right part of the matrix is the transpose of the covariance matrix among x and y. Each element in the model implied variance/covariance matrix is a function of model parameters. For a set of specific model parameters from the eight SEM fundamental matrices that constitute a SEM model, there is one and only one corresponding model implied variance/covariance matrix (Hayduk, 1987).
18.224.61.19