In Mplus Version 7.1 New Examples on slide 80 http://www.statmodel.com/examples/V7_1examples.shtml the means were constrain to zero in the reference group and freely estimated on the others. Will like to ask if that is the normal way in testing scalar invariance. do you check if the model is acceptable before the means are computed or you can introduce the means as you check the model invariance.
I have three groupings( ceod, single-sex-boy, single-sex girls). I am doing multigroup analysis. is it possible to do a pairwise analysis by way of varying the reference group and doing three analysis(three scalar models with different reference group per model). Any advice for possible pairwise comparison.