Maja Cambry posted on Friday, February 12, 2010 - 8:12 am
I'm trying to test measurement invariance across two groups in a 1-factor CFA with 5 categorical indicators. Following recommendations in 2007 user's manual I estimated the comparison model - thresholds and factor loadings free, scale factors @1 in both groups, factor mean @0 in both groups and nested model - MPlus default. Note: TYPE=COMPLEX and default estimator WLSMV was used.
Is it appropriate to use the S-B scaled chi-square for diff testing? If so, how do I get scaled chi-square values or scaling correction factors (neither is given in output)? If S-B isn't right, what method can be used for diff testing in models with categorical outcomes?