topic Re: odds ratio output from proc genmod in New SAS User

odds ratio output from proc genmod

NahidSultanakmc — Tue, 24 Sep 2024 01:11:46 GMT

I am using proc genmod for logistic regression with repeated measures. However, the following code does not show the odds ratio. I need to know the odds ratio of the continuous variable "rcg00017033 ".

proc genmod data=spt_three_final;
class studyid ;
model spt(event='1')= rcg00017033 gender P10_Q5BirthWeight/ dist=bin link=logit;
repeated subject=studyid / type=cs;run;

Re: odds ratio output from proc genmod

Ksharp — Tue, 24 Sep 2024 02:49:22 GMT

https://support.sas.com/kb/35/189.html

You need to write ESTIMATE statement for continuous variable:

proc genmod data=uti;
freq count;
class diagnosis treatment;
model response = diagnosis treatment diagnosis*treatment / dist=binomial;
estimate 'trt A vs C in comp' treatment 1 0 -1
diagnosis*treatment 1 0 -1 0 0 0 / exp;
run;

Re: odds ratio output from proc genmod

NahidSultanakmc — Tue, 24 Sep 2024 03:02:08 GMT

Thank you very much for the help.
But I don't know the use of the numbers (0,1-1 ) in the estimate statement. Is there any specific rules?

Re: odds ratio output from proc genmod

Ksharp — Tue, 24 Sep 2024 08:19:58 GMT

It is for category variable ,not continuous variable.
estimate 'age at 7' intercept 1 age 7 /exp;

Re: odds ratio output from proc genmod

NahidSultanakmc — Tue, 24 Sep 2024 18:20:27 GMT

Thank you very much again. The link is very helpful. So, if I need to know the odds ratio for two continuous variables (cg27664530, cg27664530*Time), should I use the below code?
proc genmod data=spt_three_final1;
class studyid ;
model spt(event='1')= cg27664530 cg27664530*time gender P10_Q5BirthWeight time/ dist=bin link=logit;
repeated subject=studyid / type=cs;
estimate 'cg27664530' cg27664530 2/exp;
estimate 'cg27664530*time' cg27664530*time 3/exp;
run;

Re: odds ratio output from proc genmod

Ksharp — Wed, 25 Sep 2024 00:32:42 GMT

I think you did not check URL I gave you .
https://support.sas.com/kb/35/189.html
Check example in it:
/* Odds ratio for change in StartVert from 5 to 6 */
contrast 'logOdds@x=5' intercept 1 startvert 5 startvert*startvert 25 / estimate=both;
contrast 'logOdds@x=6' intercept 1 startvert 6 startvert*startvert 36 / estimate=both;
contrast 'logOdds@6-logOdds@5' startvert 1 startvert*startvert 11 / estimate=both;

And better post it at State Forum :
https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures
and calling out @StatDave @ jiltao @SteveDenham .
I am NOT expert about this though.

Re: odds ratio output from proc genmod

Ksharp — Wed, 25 Sep 2024 00:34:44 GMT

proc logistic data=kyphosis;
model Kyphosis(event="1") = StartVert|StartVert;
estimate '4.5' intercept 1 startvert 4.5 startvert*startvert 20.25,
'5.5' intercept 1 startvert 5.5 startvert*startvert 30.25 / e;
estimate '9.5' intercept 1 startvert 9.5 startvert*startvert 90.25,
'10.5' intercept 1 startvert 10.5 startvert*startvert 110.25 / e;
ods output coef=c;
store log;
run;