02-21-2012 04:31 PM
I would really appreciate it if you could verify that I am running this mutilevel logistic regression using PROC GLIMMIX correctly.
The dependent (w) and independent (x, y, z) variables are all binary. SecondLevelID is the categorical measure of the # of sites (500 unique sites) with an average n of 45 per site. Am I accounting for the hierarchical structure of the data correctly by allowing the interecept to be random across the different sites? I am unable to calculate a c stasticis using this syntax. Am I using the correct ddfm? What criteria should I use to gauge the fit of the regression? Pearson close to 1.00?
proc glimmix data=a
model w (event=('1')= x y z /link=logit
dist=binary ddfm=bw solution oddsratio;