topic Re: ODS Output - CLODDS - not working with categorical exposure in Statistical Procedures

ODS Output - CLODDS - not working with categorical exposure

bchronister — Wed, 23 Mar 2022 04:52:09 GMT

I am trying to limit my output from my proc logistics to just my exposure of interest. My exposure, c_par_ln_q2, is a two category variable. When I use ods output to output the cloddswald, it outputs like this.

However, when I add the where statement as follows, i get a clodds1 dataset with no observations. I am not sure how to fix this.

proc logistic data=logalla desc;
by c_gender;
class c_par_ln_q2 (ref="0");
model c_testosterone_ln_q3 (ref="1")=c_par_ln_q2 c_gender /rl link=glogit;
ods output parameterestimates=pe1 (keep=variable response ProbChiSq c_gender where=(variable= "c_par_ln_q2")) CLOddsWald=clodds1 (where=(effect= "c_par_ln_q2 1 vs 0"));
run;

Re: ODS Output - CLODDS - not working with categorical exposure

StatDave — Wed, 23 Mar 2022 22:19:31 GMT

See "ODS Table Names" in the Details section of the procedure documentation whenever you have a problem like this. Since the RL option is the same as CLODDS=WALD, that table in the PROC LOGISTIC documentation indicates that the appropriate table name is CLOddsWald.

Re: ODS Output - CLODDS - not working with categorical exposure

bchronister — Wed, 23 Mar 2022 22:26:15 GMT

For the code that I provided, I did use the ClOddsWald table name. I was also able to subset it to other variables, but just couldn't specifically for the c_par_ln_q2 variable. Is there a part of the code that I needed to change, or am not seeing?

Re: ODS Output - CLODDS - not working with categorical exposure

Rick_SAS — Thu, 24 Mar 2022 13:12:57 GMT

I suspect there are TWO blanks after the variable name. Try

where=(effect= "c_par_ln_q2  1 vs 0")

Unfortunately, HTML compresses blanks, so when you cut and paste in the HTML destination, the number of blanks in the HTML might not match the number of blanks in the data string.