topic Re: How can I generate a table showing all reference categories? in SAS Programming

How can I generate a table showing all reference categories?

khalidamin — Thu, 15 Sep 2022 20:05:04 GMT

See the attached file. Thank you in advance.

Re: How can I generate a table showing all reference categories?

Reeza — Thu, 15 Sep 2022 20:20:54 GMT

Not sure what your question is, I see output from proc logistic...

Re: How can I generate a table showing all reference categories?

khalidamin — Thu, 15 Sep 2022 20:23:32 GMT

I want to generate a table as shown is the attached file using SAS

Re: How can I generate a table showing all reference categories?

Reeza — Thu, 15 Sep 2022 20:45:07 GMT

But the point estimate isn't 1, it's 0.

Re: How can I generate a table showing all reference categories?

khalidamin — Thu, 15 Sep 2022 20:49:31 GMT

Oh yes, it’s zero. Sorry for the typo.
Is there any option in proc logistic to display reference category?

Re: How can I generate a table showing all reference categories?

Reeza — Thu, 15 Sep 2022 20:57:32 GMT

Not as far as I know, you'll have to manually add those rows in, especially as the reference level can vary depending on the code so there's not an easy way to automate that ... at least that I can think of at the moment.

Re: How can I generate a table showing all reference categories?

khalidamin — Thu, 15 Sep 2022 21:43:05 GMT

Thanks Reeza for your time - I really appreciate.

Re: How can I generate a table showing all reference categories?

FreelanceReinh — Sat, 17 Sep 2022 14:14:30 GMT

Hello @khalidamin,

If you just need a dataset containing class variable names and their reference categories (which you could then "interleave" with an odds ratio table where the reference categories are omitted), you can use the ClassLevelInfo ODS output dataset:

ods output ClassLevelInfo=cli;
proc logistic ...;
...
run;

data reflevels(keep=clvar value);
set cli;
if class ne ' ' then clvar=class;
if max(of x:)=0;
retain clvar;
run;

Alternatively, you could start with the OddsRatios ODS output dataset. But unfortunately variable Effect in this dataset contains concatenated strings like 'gender 1 vs 0' so that you would need to work with character functions such as SCAN to extract variable names and values. This might become difficult in general, e.g., if values (or even variable names) contain blanks or the substring "vs", etc.