Continuous and categorical indicators PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Filipa de Castro posted on Monday, July 08, 2013 - 4:47 pm
Apologies if this is a too simple question, or previously dealt with, but I could not find it in previous threads.

I am running a LCA model to identify subgroups of patients based on a number of symptoms. The variables for symptoms are continuous, but they can also be categorized into 4-level categorical variables each level being above a threshold for severity of symptoms. There are 10 symptoms and I have a very small sample of just above 60 patients.
The model seems to work much better when I use categorical variables. It fits well into a (clinically sound) 2 class solution as indicated by several fit indices and likelohood tests. When I use variables as continuous the model never fits and all k+1 classes solution is always better than k classes.

My questions are:
1)Why is this happening?
2) when reporting this in a manuscript, how can I justify for using the categorical vs the coninuous variables?

Many thanks in advance.
 Linda K. Muthen posted on Tuesday, July 09, 2013 - 2:44 pm
When you categorize the variable, you lose power. This could cause the fit to look better. You should have a substantive reason for categorizing the variable.
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