AIC vs. BIC and LCA PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Wesley Anderson posted on Thursday, January 08, 2015 - 2:34 pm
In Nylund et al. (2007) on deciding the number of classes in mixture models, the results indicate that, in general, BIC is better than AIC. However, I am wondering if anyone can interpret why AIC performs better than BIC when the LCA has categorical outcomes, small sample size, and unequal class sizes.
 Bengt O. Muthen posted on Thursday, January 08, 2015 - 7:08 pm
Not sure. Anyone? Somehow it must relate to the fact that AIC does not punish models with many parameters as much as BIC does.
 Wesley Anderson posted on Thursday, January 08, 2015 - 8:15 pm
I suppose I'd also like to know why BIC performs better than AIC when the LCA has continuous outcomes and why BIC performs better than AIC when the LCA has categorical outcomes, small sample size, and equal class sizes too. Thanks!
 Bengt O. Muthen posted on Friday, January 09, 2015 - 4:53 pm
Not sure if anyone knows that.
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: