The model I am working on is a bivariate latent difference score model in which the difference between two time points in latent variable G (Gdiff) is regressed on the difference between two time points in latent variable D (Ddiff). Both variables' latent differences are formulated in the typical way -- i.e., by constraining the paths between the two time points to 1 and defining a difference (here Ddiff and Gdiff) as the disturbance in the corresponding variable (here D and G) at the second time point. I will call the regression of Gdiff on Ddiff "GonD."
My question is: given how the difference variables are defined, is it better-advised to interpret the unstandardized value of GonD, or the standardized value of GonD in the STDYX output? The unstandardized and standardized values of GonD are different, but given that STDYX standardizes the observed variables, it's unclear how this affects the value of GonD.