LPA with non-normal items PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Kelly DeMartini posted on Tuesday, April 08, 2014 - 12:37 pm
Good Afternoon,

I am trying to conduct a latent profile analysis using the items of a scale as the 10 indicator variables. Some of these items are zero-inflated; others are normally distributed. To run a zero-inflated model, is the only addition to my syntax the COUNT statement:

AUXILIARY = bya (e) typ (e) drug (e) h1 (e);
COUNT = slp1-slp10;
classes = slp(2);
type = mixture;

If only some of the indicators are zero-inflated, is there a different strategy that would be more beneficial?

 Linda K. Muthen posted on Wednesday, April 09, 2014 - 10:45 am
Are you variables counts? How exactly are they measurement.
 Kelly DeMartini posted on Wednesday, April 09, 2014 - 12:38 pm
The variables are on a 5-point scale with "Never" as one of the anchors, and "Always" on the other anchor. A few of the items have a typical zero-inflated distribution, with a large stack on the "Never."
 Linda K. Muthen posted on Wednesday, April 09, 2014 - 2:54 pm
You should treat this variable as a categorical variable using the CATEGORICAL option. Categorical data methodology can handle floor or ceiling effects.
 Kelly DeMartini posted on Thursday, April 10, 2014 - 1:24 pm
In an LPA model, can some indicators be categorical and some continuous? Or is it better to just treat all the items on the scale as categorical if a couple of them need to be modeled as categorical to handle floor effects?
 Bengt O. Muthen posted on Friday, April 11, 2014 - 6:11 am
You can have a blend of indicator types: binary, ordinal, continuous, count. Plus you can also have indicators that are latent variables as we show in one of our course handouts.
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