Count Variables as Moderators PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Seth J. Schwartz posted on Friday, August 16, 2013 - 1:41 pm
Linda and Bengt:

My colleague and I are running some analyses of potential moderators of treatment impact. Some of these moderators are symptom counts (e.g., total number of conduct disorder symptoms). How would we model these as moderators, and how would we explore the moderating effects that are significant? The standard deviations for these variables are greater than the means, so the Aiken & West technique of M-1SD and M+1SD will not work.

Any advice you have would be greatly appreciated.

All the best,

Seth Schwartz
 Linda K. Muthen posted on Friday, August 16, 2013 - 6:19 pm
I don't see why this would not work with M-1SD and M+1SD. I would treat the variable as continuous as it would be treated as a covariate.
 Seth J. Schwartz posted on Friday, August 16, 2013 - 7:01 pm
The only problem is that, if the SD is larger than the mean, M-1SD is a negative number. And M+1SD may be larger than any of the scores in the dataset.

If this problem does occur, how would you proceed?
 Linda K. Muthen posted on Friday, August 16, 2013 - 7:20 pm
Just evaluate it a 0, 1, and 2 etc. because it is only integers.
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: