yawen posted on Sunday, February 21, 2010 - 11:03 pm
Dear Dr. Muthen,
I have a question about non-recursive model in latent class analysis.C is a latent variable with 3 nominal indicators: A1 A2 A3.U228 a dichotomous variable and one of predictors of C. Could you please let me know how I should fix my syntax?
"u ON c" is not used in Mplus. Instead, the thresholds of u vary across the c classes (by default).
But I think this model runs into a reciprocal interaction problem. I don't understand how a variable such as u228 can both influence the class membership probabilities and have its probability vary as a function of classes.
If c is a function of u228, the u228 proportions will be different in the different c classes.