I have designed a multilevel logistic model using PROC GLIMMIX in SAS 9.3 for hierachical data based on pupil attainment where level-two is the school the pupil attends.
I'm quite sure that my model is a good fit, I have a generalized chi-square/DF value of 1.07 but have no other way of validating the model.
Are there any other methods of checking that the model is a good fit? I have read a lot about possibly extending the ROC curve to multilevel models but believe this to be invalid. Is there any multilevel equivalent of the Hosmer & Lemeshow test statistic used for single level logistic regression?