Interactions in zero-inflated poisson... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
 Nicki posted on Tuesday, May 23, 2006 - 6:38 pm
First, I must thank you all for your tremendous efforts at providing free and extensive access to your teachings--they have substatially enhanced my knowledge of these subjects and your program.

Second, I have been struggling to find examples of the kind of model I am attempting to test. I have downloaded several articles from your site and examined the Mplus examples, but I don't seem to find "exactly" what I am looking for. Would you please direct me to an article or example that might help me better understand how to:
examine interactions among continuous observed predictors of zero-inflated count outcomes over time? The interpretation of LGCM in ZIP models is difficult already (for me), but I am really stuggling to understand how I might examine/interpret interactions in this context.
Thank you so much for your time and effort,
 Linda K. Muthen posted on Wednesday, May 24, 2006 - 8:22 am
The meaning of an interaction between two continuous observed predictors would be the same as in regular regression. This does not change because of the scale of the dependent variable.
 Nicki posted on Wednesday, May 24, 2006 - 8:39 am
Thank you for your response. Do you know of any references I might consult to see how significant interactions in the prediction of level, growth, level#1, and growth#1 might be described and discussed? I would greatly appreciate some materials to reference. My outcomes are delinquency and substance use (very zero-inflated in this sample). Thank you again for your time.
 Linda K. Muthen posted on Wednesday, May 24, 2006 - 10:32 am
I think Patrick Curran has written about interactions and growth models. I don't know exact references. Aiken and West also do work with interactions. Most regression textbooks discuss interactions.

To understand what main effects mean with count variables, see:

Long, S. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks: Sage.

Once you get a feeling for main effects for count variables, interactions follow. The concept of an interaction does not differ depending on the scale of the dependent variable.
 Nicki posted on Wednesday, May 24, 2006 - 2:47 pm
Thank you, Linda. I have the Curran and Aiken & West references (I've done interactions in regression). I'll look at the Long one as well. I think I just need to spend some time with it re: categorical variables--somehow in the ZIP terms I get lost. I really appreciate you assuring me that the concepts are the same. Thank you,
 Jennie Jester posted on Monday, March 17, 2014 - 12:45 pm
Would it be advisable to center continuous variables before creating interaction terms for ZIP models?
 Linda K. Muthen posted on Tuesday, March 18, 2014 - 1:38 pm
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