ZIP GMM with 1 zero class PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Eric Wan posted on Sunday, June 01, 2014 - 7:10 pm
Hello, I would like ask a few questions about the growth mixture with zero-inflated Poisson model and set a zero class.

(1) According to the paper (Kreuter 2007), the zero class setting in Mplus is shown as below,
%c#3%
! zero class throughout
[i-q@0];
[ii-qi@0];
i@0;
[cage10-cage30@-15];
[cage10#1-cage30#1@-15];
Why do the intercepts set to be -15? Moreover, can we set the class that the score <5 instead of zero class?

(2) In Mplus, can we set the minimum sample size for each class? For example, at least 5% of the total sample is classified in each class. If not, can we set the priority for splitting new class group? For example, there are two class which class 1 and 2 have 2000 and 100 respectively. However, Mplus sometimes split class 2 instead of class 1 when classifying 3 classes analysis. So, can we set the priority to split class 1 instead of class 2 in Mplus?

Thank you very much.
 Bengt O. Muthen posted on Monday, June 02, 2014 - 5:15 pm
(1)
Q1. -15 refers to a very low logit value which corresponds to a probability close to zero.

Q2. I would not recommend trying this unless you are already and expert mixture modeler. Also, 5 is a high value when it comes to counts.

(2)
Q1. No - and I don't think you want to.

Q2-Q3. No. Mplus chooses classes that produce the highest loglikelihood; it doesn't split classes even though it might sometimes look like it.
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: