LCA with tech11 and tech14 output PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Rongjin Guan posted on Wednesday, October 21, 2015 - 9:14 am

I am running a LCA to group objects (cases) into groups. below is my syntax:

Title: my title

File is xxxx.dat;

names are var1 ... var9;
Missing are all (999) ;
usevariables = var1 ... var9;
classes = c(3);

OPTSEED = 991329;
LRTSTARTS = 0 0 100 20;

file is lca1_save.txt ;
save is cprob;
format is free;

tech11 tech14 ;

I changed number of classed, and got the following:
#classes BIC INT p-tech11 p-tech14
2 classes 63934 0.889 0 0
3 classes 63908 0.817 0.059 0
4 classes 62697 0.818 0.002 0
5 classes 62420 0.807 0.342 0
6 classes 62310 0.819 0.264 0
7 classes 62381 0.833 0.263 0(warning)

we are expecting 3 classes, but from the BIC and Intropy it is not very clear. The p-value from tech11 output may suggest a 2-class or 3-class solution, but the p-value from tech14 is always 0.

Many people said using BIC to determine the number of classes. But in this case the BIC is getting lower and lower when increase the number of classes.

This is my first time using Mplus. Would you please confirm my syntax is correct and give me some hints?

Thank you
 Bengt O. Muthen posted on Wednesday, October 21, 2015 - 3:41 pm
Perhaps there are some residual correlations within class that need to be added for BIC to find a minimum.
 Rongjin Guan posted on Wednesday, October 21, 2015 - 6:43 pm
Hi Bengt,

Do you mind telling me a little bit more how to do this? I mean adding the residual correlation within class.

Thank you

 Bengt O. Muthen posted on Thursday, October 22, 2015 - 2:07 pm
E.g. for 2 variables:

y1 WITH y2;
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