MODEL INDIRECT is not available with TYPE=IMPUTATION. You would need to use MODEL CONSTRAINT.
Liu Yue posted on Wednesday, May 09, 2018 - 6:03 pm
Hi: I have a binary data nested within person and item, so i use cross-classified modeling. Then i add some covarites for person and item respectively. How can i compute the % variance explained by each of the covarites? (if type= random) Thank you!
The percentage variance explained doesn't come out one covariate at a time (unless they are independent) it is for the whole set. The quickest way to do this is to get the factor score for the person then use a single level analysis model run where you use the same model as on the between person level, fix all the parameter estimates to those from the cross classified run and use "output:stand;" to get the R2.
Hi I have a binary data nested within person and item, so i use cross-classified modeling. Then i add some covarites for person and item respectively.Then, i want to compute the effect size for each of the covarites? How can i do that?
You would have to use plausible values for the random parameters and then compute it as in regular IRT. See http://www.statmodel.com/download/Plausible.pdf and see User's guide example 11.7 for how to get the Bayes factor scores/plausible values and page 838 from the User's Guide.
The only change I can think of is to get the full SD of the outcome going into the denominator of the effect. With cross-classified you need to add up the variances on within, between subject, and between time. That is straightforward if it is a random intercept only model.