topic Re: calculating the OR for Categorical Variables in Statistical Procedures

calculating the OR for Categorical Variables

bathshebat — Thu, 07 Apr 2016 23:29:45 GMT

How do you write a code in SAS that will enable you to calculate the odds ratio (OR) for a categorical variable such as BMI?

Re: calculating the OR for Categorical Variables

Reeza — Thu, 07 Apr 2016 23:40:28 GMT

Depends a bit on what you're trying to do, but either proc freq or proc logistic are good starting points.

If if you need more help, please post some sample data and explain what you're trying to calculate.

The documentation has good examples and example data if you can't provide your actual data.

Re: calculating the OR for Categorical Variables

bathshebat — Fri, 08 Apr 2016 02:29:03 GMT

Hi Reeza,

Thank you for your reply. I have a data set containing a variable called BMI (a categorical variable with 3 levels: normal, overweight and obese) and a variable called Exercise (a categorical variable with 3 levels: not answered, yes, and no). I want to determine which BMI group has the highest odds of not exercising. I am trying to compare the odds ratios for each group but the codes I have tried to run aren't producing the odds ratio. I have tried the following code to no avail:

data=answer2;

set answer;

if Exercise = 3;

run;

proc freq data=answer;

tables Exercise * BMI / OR;

run;

Re: calculating the OR for Categorical Variables

Ksharp — Fri, 08 Apr 2016 02:55:19 GMT

That is because your contingency table is 3x3 not 2x2 .

Try to use General Logistical Regression .

Re: calculating the OR for Categorical Variables

Reeza — Fri, 08 Apr 2016 03:28:35 GMT

Your dataset only has one value for Exercise so you have 2X1 table?

You should really post sample data.

Here's a worked through example of logistic regression that is really similar to what you're trying to do.

http://www.ats.ucla.edu/stat/sas/dae/logit.htm

My only suggestion would be to change the model line to add the expb option.

model admit = gre gpa rank/expb;

expb provides the exponentiated value of the estimates

Re: calculating the OR for Categorical Variables

ballardw — Fri, 08 Apr 2016 15:09:24 GMT

Any chance your data is from BRFSS or similar complex structured survey? If so you'll want to use Proc SurveyLogistic to use the sampling information correctly.

Re: calculating the OR for Categorical Variables

bathshebat — Sun, 10 Apr 2016 01:01:33 GMT

Thank you Xia. I will try that.

Re: calculating the OR for Categorical Variables

bathshebat — Sun, 10 Apr 2016 01:02:40 GMT

@ballardw Yes it is BRFSS data. I will try that as well. Thank you.

Re: calculating the OR for Categorical Variables

bathshebat — Sun, 10 Apr 2016 01:04:33 GMT

@Reeza Thank you very much for your assistance. I am making progress. Lol.