## Evaluating Interaction Effects - using ODDSRATIO function

Occasional Contributor
Posts: 15

# Evaluating Interaction Effects - using ODDSRATIO function

Hello,

I'm a student and this is my first time trying to create odds ratios for interactions using the ODDSRATIO function. I'm looking at complications after two procedures (variable procedure 1=A, 0=B) and I'm trying to convert back to Odds ratios after looking at the interaction of procedure with race (procedure *race). I want to create following table  where procedure=B and race=1 is the reference group. I want to compare all other odds to the odds when procedure =B, race=1.

 Odds Ratios Race A B 1 1.00 2 0.75 3 0.60 4 1.20 5 1.23

The problem is that using the ODDSRATIO column, I know only how to compare within a category (A or B) , such as

ODDSRATIO race / at (procedure = "0") or ODDSRATIO race / at (prsion, ocedure = "1). How can I compare all the races within procedure A to the reference group of procedure B (where race=1)?

Thank you so much,

Louise

Super User
Posts: 13,583

## Re: Evaluating Interaction Effects - using ODDSRATIO function

There's more than one way to create Odds Ratios. Show how you are creating yours.

It may be that another procedure would be better for your specific need but I'm not going to guess without some more details.

Occasional Contributor
Posts: 15

## Re: Evaluating Interaction Effects - using ODDSRATIO function

Hello,

Thank you very much. My current logistic regression is the following. I have left out the other variables in my model so that's it's easier to focus on just the ones with the interaction effects.

proc logistic data=data;

class procedure (ref="0") race (ref="1")  (plus other variables to adjust) /param=ref;

model tot_comp_result(event='1')= procedure procedure*race race (plus other variables to adjust);

ODDSRATIO race;

run;

My answer gives me all OR for all race combinations at procedure = A and then all OR for all race combinations at procedure = B. Is there a way to compare say the OR of (Race=2 and Procedure = A) with (Race=1 and procedure =B) using ODDSRATIO or another function? Or is it better for me to make an indicator function for all combinations of procedure x race (10 total in this case) and bypass any other functions?

Thank you.

Discussion stats
• 2 replies
• 150 views
• 0 likes
• 2 in conversation