Comparing LCGA models PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Herve Tissot posted on Tuesday, March 26, 2019 - 9:03 am
Hi,

when comparing LCGA models with different number of classes, when it comes to choose the best model, which indices should I base my decision on? I saw in other threads that entropy is not necessarily the best index to take into account. Can I rely on BIC?
 Bengt O. Muthen posted on Tuesday, March 26, 2019 - 5:16 pm
I would rely on BIC.
 Herve Tissot posted on Tuesday, March 26, 2019 - 11:49 pm
Ok thanks.
 Herve Tissot posted on Wednesday, March 27, 2019 - 1:36 am
I also saw that I could use TECH11 and TECH14 to compare models with models with k-1 class. How do you decide which model to choose when TECH11 and TECH14 have different conclusions (TECH11 with p value > .05 and TECH 14 witzh p < .05)?
 Bengt O. Muthen posted on Wednesday, March 27, 2019 - 4:13 pm
I would use BIC in that case.
 Thomas Kirsh posted on Thursday, March 28, 2019 - 1:10 pm
Besides BIC, are there any other metrics we can use to compare models? I'm trying to compare LCGA models where one has a linear transformation on the x-axis and another has a square-root transformation on the x-axis.
 Bengt O. Muthen posted on Thursday, March 28, 2019 - 5:31 pm
If you have categorical outcomes, you can use TECH10 to check bivariate fit.
 Thomas Kirsh posted on Friday, March 29, 2019 - 6:17 am
No, my outcomes are ordinal or continuous.
 Bengt O. Muthen posted on Friday, March 29, 2019 - 4:01 pm
You can look at Residuals requested in the Output command.
 Thomas Kirsh posted on Friday, April 05, 2019 - 9:47 am
Just to clarify, in order to determine if the linear or square-root transform work better, we need to compare the summed residual between the two models, right? And the residuals for each class in each model are given by the OUTPUT RESIDUAL command?
 Bengt O. Muthen posted on Saturday, April 06, 2019 - 12:48 pm
Right.
 Thomas Kirsh posted on Tuesday, April 09, 2019 - 10:12 am
Would it be okay to use both BIC and RSS to compare models in this case?
 Bengt O. Muthen posted on Tuesday, April 09, 2019 - 5:27 pm
Yes.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: