Hello, Thank you for your input. So I just tried creating a tiny dummy dataset that does not have all of my variables nor my observations, but it includes the one variable (FSA) that I would like to change. This is what I wrote: data test; input FSA$ agesex hypertension ischemicheartdisease heartfailure; datalines; B0A 1 0 0 1 B0A 5 0 0 0 B0C 7 0 0 0 B0C 4 0 1 0 B1J 3 0 1 0 B1J 9 1 0 0 B1J 2 0 0 0 ; So FSA(area variable) is the only character variable, whereas agesex (10 age and sex groups) and hypertension (binary), ischemicheartdisease (binary) and heartfailure (binary) are all numeric variables. I would like to convert the FSA variable to numeric as well, but retain the observations as written in the rows (so B0A, B0C, B0C, etc.). Is this even possible using that instruction file with the link that I posted? I tried rerunning it on this example dataset, and although I did not experience any errors in the input function statement, the resulting dataset had missing observations '.' for all FSA observations. Alternatively, I was wondering like I mentioned before, the reason I am trying to do this is because I am having trouble running a proc glimmix model. The FSA variable is specified in several steps of the regular model statements: proc glimmix data=have; class agesex FSA; model ischemicheartdisease(event="1")=FSA / dist=binary link=logit s; lsmeans FSA/ cl ilink; random in/ subject=FSA; run; The model does not converge. I was thinking that mabye this is because the FSA variable must be a numeric variable for this model to run, but is this actually true, and if not, can I first ignore all of what I am trying to accomplish up top? And then instead on figuring out what I may be misspecifying in my proc glimmix model and why it is not converging? Any possible ideas on what it does not converge and what I would need to share to try to diagnose this? Thanks again, Bart
... View more
