I am using DCON with a distal outcome. I wish to include two other covariates in the model - can you advise as to which Auxiliary command to use or should these covariates be specified in the model command?
Also I note the error message that DCON cannot be used with KNOWNCLASS - is there a workaround this?
I used auxiliary command and dcat/dcon to compare frequencies and proportions of distal outcome variables. However, the result was far from expected. The results of auxiliary variables were not consistent with cross-tabulations and mean comparisons using most likely class membership from the same input program. The entropy was .885, so I don't think the uncertainty came into play. One thing I'm suspicious was the class order switched. There were three classes, in the descending order of 1-2-3 in terms of class size. However, in the distal outcome section of output, the classes were in the ascending order of 3-1-2 in terms of approximate SE. In addition, if I switched the order by its approximate SE, the result was consistent with crosstab and mean comparison using most likely class membership.
1. Is it possible that classes switch in distal outcome output using dcon/dcat option? 2. If so, how can we check the correct order of class. 3. If not, could you guess what could possibly went wrong? 4. In this case, with entropy .885, can I report proportion and mean comparisons based on most likely class membership, instead of using auxiliary command? 5. Could you suggest any better ways?
To follow up on this thread - this user ran into a problem in version 7.11 where DCAT/DCON gave incorrect results when TECH11 and TECH14 were also requested. This will be corrected in version 7.3.
Dina Dajani posted on Thursday, December 03, 2015 - 11:12 am
Can you please advise on the first question of this thread:
How do you specify a covariate when performing a regression with Lanza's DCAT. I am regressing a categorical distal outcome on my latent class variable and wish to include a categorical covariate (Distal ON Covariate).
Asparouhov, T. & Muthén, B. (2014). Auxiliary variables in mixture modeling: Using the BCH method in Mplus to estimate a distal outcome model and an arbitrary second model. Web note 21.
Dina Dajani posted on Thursday, December 03, 2015 - 6:32 pm
Thank you very much Dr. Muthen,
I also have one distal outcome that is categorical. According to the Web note you posted above, DCAT is my best option. But when I used this, my results seemed very strange (see below). The probabilities look correct but the odds ratios and CI's do not. Do you know what might be the problem? Thank you for any help. Best, Dina Dajani
PRIDX Prob S.E. Odds Ratio S.E. 2.5% C.I. 97.5% C.I.