Proc genmod with categorical covariates

[zihdonv19]

I want to fit a model below, where edu is a 3-level variable, cou is a binary variable:

proc genmod data=cohort descending;
class id edu (ref='1');
model cou=edu female / dist=poisson link=log;
repeated subject=id/type=ind;
estimate 'edu 2 vs. edu 1' edu 0 1 0;
estimate 'edu 3 vs. edu 1' edu 0 0 1;
ods output Estimates=outtab;
run;

 But the results viewer showed "Non-est" in the contrast estimate results table. I found people always refer to this note (https://support.sas.com/kb/24/447.html) for relevant questions, but I felt so lost after reading this. 

I just want to know:

edu 3 vs. edu 1, RR=....95%CI=...

edu 2vs. edu 1, RR=....95%CI=...

[PaigeMiller]

Use the LSMEANS command in PROC GENMOD, with the DIFF and CL options. https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_genmod_syntax20.htm

[Ksharp · Accepted solution]

Check this:
http://support.sas.com/kb/24/188.html

I guessed you should code something like:

proc genmod data=cohort descending;
class id edu ;
model cou=edu female / dist=poisson link=log;
repeated subject=id/type=ind;
estimate 'edu 2 vs. edu 1' edu -1 1 0 /cl ;
estimate 'edu 3 vs. edu 1' edu -1 0 1;
ods output Estimates=outtab;
run;

Check :
https://support.sas.com/kb/24/447.html