Arya Ansari posted on Wednesday, August 21, 2013 - 11:59 am
I have conducted latent profile analysis (five classes) using the class membership variable. I imported these into SPSS and created a dummy coded class membership indicator.
My question is regarding the standardized estimates (STDYX/STDY) in subsequent regression models in Mplus.
I am interested in each contrast between the groups; however, when switching the referent group the standardized estimates change for the same contrasts even though the unstandardized estimates are the same. The only thing being changed between the models is the referent group. Although the differences in some standardized estimates are minor (i.e. .108 versus .102) there are ones that are larger.
When Profile 4 is the referent comparing class 4 with 2 I get: unstandardized = -2.13, standardized = - .069
Now when Profile 2 is the referent comparing class 2 with 4 I get: unstandardized = 2.13, standardized = .046