I am using PROC CALIS to model an endogenous latent variable and PROC SCORE to obtain predictions for it.
I found that the OUTSTAT dataset produced by PROC CALIS does not include intercept estimates for this latent variable. This was causing PROC SCORE to be off in its predictions. I did use the options MEANSTR in PROC CALIS and NOSTD in PROC SCORE (so those were not the causes of this issue.)
When PROC REG generates an OUTTEST dataset with the estimated model coefficients, it includes a column named “intercept”. I would suggest that similar functionality is implemented for the OUTSTAT dataset of PROC CALIS (where all the other estimated model coefficients are.)
I found that PROC CALIS does save the intercept in the OUTMODEL dataset. So, I manually wrote a data step to get that number and add it to all predictions that came from PROC SCORE. This solved the problem. But it would be desirable to not have to do this manually. It would save time. It would be safer too (because one could miss noticing this issue) specially if one is used to PROC SCORE completely handling this.