How to Model Longitudinal Data with Interval Data as Response Variable

[sabataged]

Hello, 

I'm having trouble figuring out how to model my data. I am trying to determine gender difference in childhood PTSD at 3 different timepoints. 

I have the following:

Predictor variables:  Fixed effects such as gender, neighborhood, age, and race. 

Outcome variable: PTSD Burden/SCORE

• PTSD score is determined by tallying up the score for 17 items on a questionnaire.  For each question, there are 4 ordinal choices ranging from 0-3.  The answers to these questions add up to a total of 51 possible points. This scale can be considered as interval data. 

I was modeling PTSD Burden as a continuous outcome using PROC MIXED, but after some reflection and research, I realized this might not the best way to model this data. For example, I was getting an intercept of 21.56, but it is impossible to get a non-integer PTSD score. I cannot find an integer option with PROC MIXED.  As far as I know, only PROC OPTMODEL can give integer solutions.  Do I need to transform my data in any way?  Would anyone be able to point me in the right direction?

Thanks so much! 

[Rick_SAS · Accepted solution]

Personally, I don't have a problem saying that the average PTSD score (at time=0?) is 21.56. It's an average. We routinely hear that the average American family has 2.5 kids and 1.2 pets. Averages don't have to be integers.

Moving this topic to the Statistic Community for greater visibility.