LPA with both categorical and continu... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Danielle Ostendorf posted on Monday, November 06, 2017 - 1:24 pm
I am using LPA to examine if latent profiles form from 8 indicator variables. 1 indicator variable is heavily weighted at 0. So, I categorized this variable into 4 categories. The model runs when covariances are fixed to 0 and variances are equal across classes, but I donít know how to check the different variance/covariance structures when I treat the ordinal variable as categorical.

Also, when I use a sensitivity analysis to see if I treat the ordinal variable as continuous, I get different results (evidence for 4-class model) compared to when I treat the ordinal variable as categorical (evidence for 3-class model).

Main Questions:
1. I received an error about needing theta parameterization to examine covariances for when I treat the ordinal indicator as categorical. Iím not sure if/how parameterizing the model in that way would impact findings for the model as a whole (i.e., given that most of my indicators are continuous, will parameterizing the model in this way change the parameterization for my continuous indicators?)

2. Would it be simpler/kosher if I examined class differences in the variance/covariance matrix for my 7 continuous variables but simply ignored the possibility of testing the conditional independence assumption for my 1 ordinal indicator?
 Bengt O. Muthen posted on Tuesday, November 07, 2017 - 2:25 pm
1) Theta param'n is relevant only for WLSMV which doesn't do mixtures. You can add a single factor and see if some loadings are significant, indicating non-independence conditional on classes.

2) Better to use a factor.
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