Continuous distal outcomes with LPA PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Pamela Kaliski posted on Wednesday, August 08, 2007 - 6:32 pm

I am doing a latent profile analysis, with six continuous indicators used to form the latent classes. Additionally, I have 5 continuous predictors and 1 continuous distal outcome. I cannot figure out the correct Mplus syntax for modeling a continuous distal outcomes. I looked up distal outcomes in the Mplus user's guide index, which refers me to pages 184-185. However, the example shown on these pages is for a binary distal outcome.

What is Mplus syntax is used to model a continuous distal outcome?

Thank you!
 Linda K. Muthen posted on Tuesday, August 14, 2007 - 2:11 pm
It is the same as for categorical except you don't use the CATEGORICAL option. Include the variable on the USEVARIABLES list. The relevant parameter is the mean of the distal outcomes in each class.
 Pamela Kaliski posted on Tuesday, August 21, 2007 - 8:51 am
Thank you Linda for your response. I have a follow-up question. Given that the mean is the relevant parameter of the distal outcome in each class, is there any way of testing whether or not the means of the distal outcome are significantly different between the classes? In other words, is there a way of knowing from the Mplus output if class membership is a significant predictor of the distal outcome? With the logistic regression of latent class membership on the continuous predictors in my model, there is a test of whether or not the predictors are significant predictors of latent class membership. Is there a similar test of whether or not latent class membership is a significant predictor of the distal outcome?
 Linda K. Muthen posted on Thursday, August 23, 2007 - 3:12 pm
You can use MODEL TEST to test if these means are different. This gives a Wald test. See MODEL CONSTRAINT for a description of how to label the parameters in the MODEL command. The labels are used in MODEL TEST.

Think of the regression of a continuous outcome on a binary covariate,

y = b0 +b1*x

When x=0, y=b0. When x=1, y=b0 + b1. There is simply a shift in this case. Mplus estimates b0 for one class and b0 + b1 for the other class. By testing the equality across classes, you are testing if the difference between b0 and b0 +b1 or b1 is signficant.
 Hamed Q. Saremi posted on Wednesday, March 01, 2017 - 12:24 pm

I am trying to run LPA with two continuous indicators and one continuous distal outcome, using cross-sectional data. I would like to see if the latent classes generated by the two indicators can explain any variance in the distal outcome. I looked up the Mplus guide version 7 and the only example providing a sample syntax for LPA with distal outcome seems to be Example 8.6. However, that example is for cases with longitudinal data and includes other latent and observed variables in its model. My data is cross-sectional and I do not have any other variables in my model. I was wondering how I should change the Syntax to run my analysis in Mplus.
When I removed the parts in Example 8.6 syntax that are not related to my model, what is left is essentially an LPA based on three indicators (i.e., the distal outcome is considered as an LPA indicator along with the two other continuous indicators that I have). I doubt if this is correct. I appreciate your advice on what I am missing here.

Thank you
 Bengt O. Muthen posted on Wednesday, March 01, 2017 - 5:56 pm
Do it using

Auxiliary = distal(BCH);

where "distal" is the name of your cont's distal outcome.

See our 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.
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