Anonymous posted on Thursday, April 22, 2004 - 2:10 pm
I have a binary latent class growth model with 7 classes. I would like to specify ethnicity (white, black, hispanic, and other) as a predictor of class membership. I have 3 dummy coded variables to represent ethnicity (with black subjects as my reference group). One of my classes has no white subjects. The model will not terminate. Is it permissable for me to fix the effect of being white for this particular class? If so, how should I do it?
bmuthen posted on Thursday, April 22, 2004 - 2:38 pm
Fix it to the large value that you get in the output. This says that being white implies that membership in this class is next to impossible. I think Mplus Version 3 does this automatically.
Anonymous posted on Sunday, April 25, 2004 - 7:01 pm
Thank you. I also have a follow-up question. I'd like to specify an a priori class that has a 0 probability of experiencing the event at each measurement occasion. I'd like to constrain the intercept and growth parameters to represent this. Is it appropriate to constrain the intercept to be -15 (the max threshold limit for binary outcomes if I understand correctly), and the slope and quadratic terms to be 0 for this class?
bmuthen posted on Sunday, April 25, 2004 - 7:57 pm