Tor Neilands posted on Tuesday, November 13, 2018 - 10:44 pm
We seek to fit a model of the form C -> M -> Y where C is a latent class variable with three classes based on a latent factor measured by four indicators. M is a continuous latent variable measured by 9 indicators and Y is a binary observed variable. We are having some difficulties with specifying the regressions of M and Y on C. We have tried, e.g., M ON C#1 C#2 ; but obtain an error message. We are wondering if anyone has a suggestion for how one could specify a latent class variable as an X-variable in Mplus?
M ON C is not allowed. Instead of M ON C, you would study the effect that C has on M by how the M means change over the classes (that's after all what regression with observed dummy variable does). Same for how C influences Y.
In the second run, I also want to see whether or not the observed indicators (the ones that were used to organize groups in the first run) predict a distal outcome while controlling covariates IN EACH CLASS.
Can I do that? If the answer is positive, do you mind telling me the codes?