LCA_questions PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Cecilia Reyna posted on Tuesday, March 31, 2009 - 9:25 pm
I have a few doubts about the LCA

1) How many indicator variables can be included in the model?

2) Is there a relationship that should be considered between sample size and parameters to estimate?

3) Having achieved the probability of belonging to each class of individuals, there is a cutoff point that could be used? For example, if the probability of the subject 1 is =
Class 1: 0.20
Class 2: 0.40
Class 3: 0.40
Class 4: 0.20
Should consider this case for further analysis? Is there some literature on this point?

 Bengt O. Muthen posted on Tuesday, March 31, 2009 - 9:53 pm
2) Your sample size should be considerably larger than your number of parameters. To be more precise about the sample size needed, you have to do a Monte Carlo study.

1) It is easy computationally do handle at least 100, but see answer to 2)

3) Your example is a person with poorly determined class membership. You should strive for at least 0.80 in one class. But in terms of relating the latent class variable to other variables, you don't have to use this multi-step, analyze-classify-analyze approach, but do it in one step. There will shortly be a paper posted on this topic.
 Cecilia Reyna posted on Tuesday, March 31, 2009 - 10:26 pm
Thanks Dr Muthen.
I wait for the paper.
(autocorrection: in the example (Q3), Class 1 and Class 2: 0.10)
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