Suppose I have two endogenous variables (A and B) and I draw an arrow (double-headed) between them. What is openmx doing in the background? Is it fitting a correlation between the residuals of A and B?
If they are endogenous (i.e. they are variables with one-headed arrows going in to them), then the path with a double-headed arrow between them is a residual covariance. In LISREL terminology, the two-headed arrow between these variables goes in the Psi matrix, abbreviated as PS.
That's what I thought. Thanks Mike!