09-12-2016 12:55 AM
Dear SAS programmers
I'm currently analysing relatively large dataset (n=~20,000), you may think this is not large enough. However, I've got a mixed of 12 studies, some of them are individually matched.
Anyway, I was using GLIMMIX to run the model in which I got confused when I tried to fit two random statement in the model.
Basically, I've got a variable "study" representing 12 studies, i.e. 1, 2, 3....12.
I also have a variable "matchedid" which is used for individually matched data. For studies without individually matched data, a single number was given.
The model looks like this (note, 1-6 is just the line, not the actual code in case someone misunderstands it):
1 PROC GLIMMIX DATA = A;
2 CLASS study matchedid x;
3 MODEL outcome (EVEN = '1') = x / SOLUTION DIST = binary LINK = logit ODDSRATION (label) DDFM = BW CL;
4 RANDOM int / SUBJECT = study;
5 RANDOM int / SUBJECT = matchedid (study);
When I tried to run the model by looking at the random effects of study only ommiting line 5, it's sig. for study. which means that I need to consider the variation across studies?
However, it seems like it will take SAS forever to run the whole model.....I'm just wondering if I've done something wrong here or it's just because of the size of the matrix.....or whatever reason.
My Q is that how do you incorporate multiple individually matched studies in one GLIMMIX model to account for the random effects of studies for individually matched studies?