Hello,
I am trying to measure the difference in change of eGFR (postindex_egfr) between treated (cohort_num=1) and untreated (cohort_num=0) groups. I am using a repeated measures mixed model to study this association and eGFR values have been divided by quarter (over 5 years). A lot of quarters will have missing eGFR values.
There are no duplicate values by quarter for any given patient. The code is written below and I'm getting an infinite likelihood error. Advice would be appreciated.
proc mixed data=hcvrenal.analysis_set_3 method=reml covtest;
class patid cohort_num (ref='0') gender bus_num dm(ref='0') htn(ref='0') ccf(ref='0') mi(ref='0') pvd(ref='0') cva(ref='0') can(ref='0') abuse(ref='0') hiv hbv ftg(ref='0') ckd_stage(ref='1') quarter (ref='1');
model postindex_egfr = cohort_num ageatindex gender bus_num dm htn ccf mi pvd cva can abuse hiv hbv ftg fib4 ckd_stage QUARTER indexdate /s ddfm=satterth COVB ;
WHERE QUARTER IN (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20);
RANDOM patid/cl;
repeated quarter/type=ar(1) subject=patid*cohort_num ; lsmeans cohort_num/cl;
estimate 'Change from baseline to quarter 4' quarter -1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0/e cl;
estimate 'Change from baseline to final eGFR' quarter -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1/e cl; run;
... View more