Thresholds in GMM w/ categorical dist... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Alex He posted on Sunday, February 08, 2015 - 10:06 am

I was working on Ex8.6 from the user guide. I wonder how I can interpret the thresholds associated with the distal outcome within each class. Take the output of Ex8.6 for example, for Latent Class 1, does -15 for U$1 indicate that the logit of getting the outcome "1" is -15? (Basically the odds are very very small). Does 4.194 associated with U$1 under Latent Class 2 indicate that the logit of getting a "1" is 4.194, which means the odds are very high?

A related question is why there is no significance test associated for the threshold for Latent Class 1? Is there a way in Mplus to statistically compare the thresholds of Class 1 and Class 2? Thanks!

Latent Class 1

U$1 -15.000 0.000 999.000 999.000

Latent Class 2

U$1 4.194 0.864 4.857 0.000
 Bengt O. Muthen posted on Sunday, February 08, 2015 - 12:35 pm
I don't see those parameter estimates for UG ex 8.6, but let me answer in generalities. A threshold of -15 means that that the lowest category of the observed item has zero probability of occurring. A high logit of 4.194 implies a high probability for the lowest category.

You can test threshold equality by using Model Test,

0 = t1-t2;

where the t's are labels for the thresholds given in the Model command.
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