milan lee posted on Tuesday, March 24, 2015 - 5:29 am
Dear Dr. Muthen, I read the Mplus online workshop slides ( IMPS 2014, Madison) in which the method for testing measurement invariance was presented. On Slide68, it was mentioned that metric invariance is needed to compare factor variances across groups whereas scalar is mainly for comparing factor means. Can I interpret it as if I am comparing invariance of structural path coefficients across countries, measurement invariance built only at the metric level is okay? When I was trying to conduct measurement invariance across two countries for 6 measures, I could establish metric (factor loadings) invariance only but failed to do so for the partial or full scalar invariance. May I have your advice on relevant references that I can cite? Thank you!
Q2. You can try partial scalar invariance, where you let some offending intercepts/thresholds be non-invariant. Partial invariance is discussed in the Byrne, Shavelson, Muthen article and also in our short course teaching - see Topic and 2.
milan lee posted on Tuesday, March 24, 2015 - 2:50 pm
Dear Bengt, Thank you for your advice! Just wanted to clarify my previous question, for the 6 measures I used across countries, I could not make partial scalar invariance achieved. So, I established invariance only at the factor loadings and then compared invariance of path coefficients. Are there any references you would recommend to endorse my attempt to test structural path differences based on metric-only measurement invariance? Thank you!