GMM and auxiliary variables PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
 Eivor Fredriksen posted on Wednesday, October 21, 2015 - 4:19 am

I'm doing a GMM resulting in a 4-class solution. I now want to predict classes (the latent categorical variable) using auxiliary variables by using the 3-step approach (AUXILIARY = var (R3STEP);). My question concerns whether I can obtain odds ratios and confidence intervals for the regression part of the model? And how to do that?

Thank you!
 Tihomir Asparouhov posted on Wednesday, October 21, 2015 - 5:58 pm
Neither of this are in the output but you can compute them from the Mplus results.

Odss Ratio = EXP(regression parameter estimate)

Confidence Limits = regression parameter estimate +- 1.96 * (its standard error)
 Tihomir Asparouhov posted on Thursday, October 22, 2015 - 4:48 pm
To get the confidence limits for the Odds Ratio just exponentiate the above confidence limits for the regression parameter.
 Biyao Wang posted on Thursday, December 03, 2015 - 2:05 am
Dear Tihomir Asparouhov,

I'm doing a GMM using 3 wave data and resulted in a 4-class solution. I want to test the relationship between some wave 1 factors and the latent class. The classification was not very high (entropy=0.67) so I want to use the 3 step approach.

(1)to test the predictive value of factors, I used R3STEP so the wave 1 factors are used as predictors of latent class.

(2)to test the equality of means of factors across latent class, what should I do?
Can I use DU3STEP or DE3STEP? What is the difference between them?
Can I use wave 1 factors? (in your 2014 paper, you said in DU/DE3STEP variable is distal variable, what is a distal variable? is it the same with distal outcome? does it mean the variable should be assessed at wave 3?)

Thank you!
 Bengt O. Muthen posted on Thursday, December 03, 2015 - 6:49 am
(1) you would have to use a manual version of R3STEP to let a factor predict class because R3STEP only allows observed predictors.

(2) Use BCH - see the paper on our website:

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.

This paper has tables at the end describing differences between approaches.

I don't think it matters when the auxiliary variable was measured.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message