04-25-2018 01:25 AM
Hello all. Hoping you can help. My dataset has 850 observations over 16 variables. I am using SAS 9.4 through my university. I am trying to model a continuous outcome of Grade Point Average that is predicted (hopefully) by ordinal data. The issue I'm running into is that my continuous outcome has 3 repeated measures: GPA at end of 8th, 9th, and 10th grades. The data is currently arranged in wide format, like this:
SubjectID CovariateX CovariateW CovariateY ... ... Gpa8 Gpa9 Gpa10
I am hypothesizing that the trend GPA follows from grade 8 to 10 is predicated on the ordinal covariates X, W, Y, etc.
However, the GPA repeated measures are coded uniquely so I am not sure how to create my MODEL statement.
I've tried the following:
-create 3 different data sets from parent dataset, omitting the 2 GPA variables I don't need. Recode the GPA variable to "GPA". Merge the 3 datasets by SubjID hoping that GPA coding stays in place for all 3 GPA variables (8, 9, and 10). Then I could MODEL GPA=covariateX|covariateW|covariateY
-MODEL gpa10-gpa8 under PROC GLM, but something didn't look right and it doesn't make sense in my head. Maybe this is the correct way to do it? What about PROC MIXED? MIXED didn't like the '-' symbol in my model statement.
-Creating a new dataset and renaming gpa8, gpa9, and gpa10 to 'gpa' to indicate a common continuous response. SAS did not like that.
So to summarize, I need to model a continuous outcome of GPA based on several ordinal variables. However, my repeated GPA responses are currently contained in 3 different, uniquely-coded columns.
I hope someone can help me. Thank you!
FYI: I've attached a screenshot of the dataset to show you what I mean by how the data is stored.