anonymous posted on Wednesday, February 22, 2006 - 7:47 pm
I am doing path analysis with observed binary and continuous variables. My estimator is WLSMV and I am using Theta parameterization. In my model I use the WITH command to correlate my mediating variables. If y1 and y2 are continuous and y3 and y4 are binary, how do I interpret the coefficients for y1 WITH y3, y3 WITH y4 and y1 WITH y2? Are all of these coefficients interpreted as correlated residuals? Or are they correlated residual variances? Or is it the observed variables that are correlated?