topic Re: Which analysis to choose? in Statistical Procedures

Which analysis to choose?

maria_sas — Mon, 24 Oct 2016 21:28:41 GMT

I´m new to SAS and statistical analyses. I hope I´m posting in the right forum.

I´m trying to perform a logistic regression on my data.

My dataset contains patients with a chronic disease and a controlgroup. They are pooled in no disease=1, light disease=2 and severe disease=3 (90 patients and 250 controls, matched by age groups). I want to investige the relation to impotence, impotent=1, not impotent=2

My statement is as follows:

proc logistic data = controlsandpatients;
class disease (ref = "1") / param=ref ;
model impotent (event = "1") = disease / lackfit ;
run;

I get a quasi-complete separation problem, The maximum likelihood estimate may not exist. I´ve tried to do an exact test, but that doesen´t solve the problem. Is this the right way to analyse this or should i use PROC FREQ?

I hope you can help me

Best regards

Maria

Re: Which analysis to choose?

Ksharp — Tue, 25 Oct 2016 05:39:17 GMT

Since you have control group, do you consider to use conditional logistic regression ?

Re: Which analysis to choose?

maria_sas — Tue, 25 Oct 2016 08:41:47 GMT

Thanks for your reply.

I am new to sas programming, and have never done a conditional logistic regression - I tried looking in sas support pages but I am not sure how I would set it up?

Re: Which analysis to choose?

Ksharp — Tue, 25 Oct 2016 09:16:26 GMT

There is an example in Documentation about conditional LR.

I remember you should use STRATA statement .

Re: Which analysis to choose?

maria_sas — Tue, 25 Oct 2016 09:18:10 GMT

Thanks!