If we fit GEE model using PROC Genmod to a binary response data using logit link and unstructured or exchangeableworkign correlation matrix, is there any way we can test for significance of the correlation we obtain?
To be more specific followign is the way my data looks like
ID
| Physician # | Patient ID | Outcome (Binary response) | Gender
| Score out of 25 (say)
|
---|
1 | 1 | 1 | 1 | M | 5 |
| 1 | 2 | 0 | F | 15 |
| ;; | ;;; |
|
|
|
| 2 | 1 | 1 | M | 10 |
| 2 | 2 | 1 | M | 24 |
| ;;;; | ;;;;; |
|
|
|
|
|
|
|
|
|
| N | 1 | 0 | F | 23 |
| N | 2 | 0 | M | 21 |
| ;;;;; | ;;; |
|
|
|
And below is the SAS code for the GEE model:
proc genmod data=XYZ;
class patientID physician gender;
model outcome=gender score / d=bin;
repeated subject=PatientID(physician) / type=unstr corrw;
run;
I would really appreciate if any one can guide me in this regard.
Thanks
Tasneem