05-03-2013 04:56 AM
Hello SAS Users!
I am working on a longitudinal data analysis ( to determine risk factors for individual infection at household level), i have adopted GEE model (proc genmod step with repeated measure option as household (Below is my SAS codes).
proc genmod data=data_gd
class sub_id hse_id;
model sub_id=V1 V2 V3 V4 V5 V6 V8;
repeated subject=hcode / logor=fullclust Type=ind;
AS we know all GEE model, we use the QIC/QICu value to determine the most robust model, not R² squared as if for logistic regression models.
Iam having difficulties to design or get the best SAS CODES for evaluating my fitted model, Can someone suggest to me the best codes for the following procedure;
(1). Assessing summary measures of goodness of fit of my GEE logistic regression model
(a) to compute the ROC curve, which is a measure of discrimination,
(b) Hosmer-Lemeshow test, which is a measure of calibration.
(2). Logistic regression diagnostics (using residuals or leverage)
Thanks in advance
Iam from non mathematical background (medical) and don't have such a strong math skills.
I have been self teaching SAS procedures, then apply on to my data for analysis.