Classes with very low numbers of subj... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Daniel Cohen posted on Friday, December 18, 2015 - 2:18 pm

I am working on a LPA and we have determined that the appropriate number of classes is 5 based on the lowest BIC. However, we have one class that only has 7 subjects (4% of the sample). Is it problematic to have such a small sample within a particular class?

 Bengt O. Muthen posted on Friday, December 18, 2015 - 5:48 pm
Mostly in terms of trying to relate that class to other variables.
 Jon Heron posted on Saturday, December 19, 2015 - 12:32 am
It's fairly common to use class size as a selection criterion. Probably not good practice from a statistical perspective but its pointless having a solution you can do nothing with.
 Daniel Cohen posted on Tuesday, January 12, 2016 - 8:50 am
Thank you so much for your responses. I'm a bit of a novice with these types of analyses, so please bear with me. Could you provide a little more detail regarding why you couldn't relate a very small class to other variables? What are the consequences of selecting a solution that includes a very small class?

Thank you!
 Bengt O. Muthen posted on Tuesday, January 12, 2016 - 6:57 pm
The parameters that are specific to that relationship would be estimated by so few subjects that their estimates and SEs might not be good.
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