Zero-inflated poisson models PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 Scott C. Roesch posted on Thursday, October 16, 2008 - 11:37 pm
Does anyone have a good example (i.e., sample MPlus code) of a zero-inflated poisson model using the TWOLEVEL RANDOM command? Thanks!
 Linda K. Muthen posted on Friday, October 17, 2008 - 2:41 pm
You can start with Example 9.2 and add:

COUNT = y (i);

and change the MODEL command as follows:

s | y ON x;
y#1 ON x;
y s ON w xm;
y WITH s;
 Scott C. Roesch posted on Saturday, October 25, 2008 - 12:15 pm
Thanks Linda! One more question if you don't mind. When running my ZIP model MPlus obviously creates the binary variable (mine is alc#1). I assumed, but wanted to confirm, that this alc#1 variable takes the 0s from my original count variable (alc) and gives them 0s in alc#1 and anyone that has a non-zero value on alc gets a 1 for alc#1.
 Bengt O. Muthen posted on Saturday, October 25, 2008 - 4:52 pm
I am quoting from my ZIP post yesterday:

"In the binary part Mplus does logistic regression for Prob(u#=1), where u# is a binary latent inflation variable and u#=1 indicates that the individual is unable to assume any value except 0. This is called the zero class. "

So anyone who has a non-zero value on alc gets a 0 for alc#1. In other words, alc# (alc#=1) represents the inflation - the 0 part.
 Stella Resko posted on Wednesday, November 19, 2008 - 10:16 pm
I was wondering if anyone has a good example (i.e., sample MPlus code) of a zero-inflated poisson model that includes an interaction? Thanks!
 Bengt O. Muthen posted on Thursday, November 20, 2008 - 12:34 am
Do you mean interaction between 2 IVs? If so, you just use Define to create that interaction and use it as another IV.
 Paige Lembeck posted on Monday, August 04, 2014 - 8:31 pm

I was wondering if there is a limit on how many variables can be treated as count variables within a ZIP model? Can only endogenous variables be zero inflated for the model to converge?


 Bengt O. Muthen posted on Monday, August 04, 2014 - 11:12 pm
Q1. No

Q2. Typically ZIP modeling is applied to DVs. After all, how would you formulate influence from a ZIP variable (that is, having the ZIP variable as an IV)? Treating the counts as a continuous variable would be unsatisfactory, I would think.
 Paige Lembeck posted on Monday, August 04, 2014 - 11:22 pm
Dr. Muthen,

Thank you for your response! That helps.


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