I am new at MPLUS and I am conducting a simple CFA with all continuous variables. Can someone tell me how I can obtain the standardize loadings and correlations as well as robust statistics? I tried to find out in the manual but I could not figure it out. Thanks
You can obtain standardized loadings by asking for STANDARDIZED in the OUTPUT command. See the OUTPUT command in the user's guide to see other optional output. I don't know what you mean by robust statistics.
Juliana posted on Thursday, March 21, 2013 - 5:55 pm
Dear Profs Muthen,
I am comparing two models, CFA and MIMIC:
M1) I am running a CFA model with 2 factors and binary indicators, testing for strong measurement invariance of the two instruments. Indices and Chi2 test indicate a bad model fit.
M2) After including one covariate, allowing a mean differences for the factors, results give a good model fit.
M3) What comes to a surprise to me: when i fix both regression coefficients to 0 again, model fit is fine, as well. But I expected M3 and M1 to be the same model... and paramter estimates are very close, indeed, through estimates of thresholds vary slightly, maybe.
Which difference explains the different model fits?
In M2,I assume you let the covariate influence the 2 factors only. And I assume that in M2, you do an invariance test across the groups.
Based on your M3 results it sounds like your covariate does not have strong influence on your factor in your M2 run. If the covariate does have a significant influence on the factor, a model test in each group (configural model) should point to rejection since the covariances between the covariate and the factor indicators are not well fitted. I would hesitate to do invariance testing when the configural model doesn't fit.