|
Message/Author |
|
Dongying Li posted on Friday, August 03, 2018 - 2:08 pm
|
|
|
I am conducting an LCA following the procedure and codes specified by Henry & Muthen, 2010. After model selection, I chose a 2-level, 3 class model with the parametric approach. However, in the model output of the 2-level model, I only got the threshold estimates, not the results in probability scale. I wonder if there is a way to get results in probability scale for the 2-level models, as in the 1-level model? If not, how should I interpret the class characteristics based solely on the threshold estimates? Thank you very much! |
|
|
This is not automatically obtained in the output. But you can compute it by hand for specific factor values, using the approach discussed in the paper on pages 204 (see last paragraph) and 206 (first paragraph). |
|
Dongying Li posted on Tuesday, August 07, 2018 - 7:26 pm
|
|
|
Thank you very much. I was able to calculate the probabilities using p=exp(estimate)/1+exp(estimate). But I am not sure how to calculate the SE. Can I calculate that based on the outputs? Also, as I understand, for the parametric multilevel approach, we are allowing the probabilities of individuals belonging to classes to vary across level 2 units; but given a certain class, the probability of responding to a certain category for an observed variable is constant, right? I apologize if the questions seem too basic. I am a new user of Mplus and know very little of latent models. Thank you very much. |
|
|
You should express your p in the Model Constraint command using parameter labels from the Model command - this gives you the SEs as well. Even within a certain class do you have variation across level 2 units. Being a beginner in mixture modeling, it is indeed hard to jump into one of the most advanced versions of mixtures. |
|
Back to top |
|
|