topic Odds ratios not showing for individual variables in interaction term? Logistic regression in Statistical Procedures

Odds ratios not showing for individual variables in interaction term? Logistic regression

christinef — Thu, 21 Apr 2022 13:17:23 GMT

Hello,

I am trying to run a multi variable logistic regression model predicting departure from position (yes/no) using perceived culture around alcohol as the main predictor with sex, age, race, hazardous alcohol use, and job satisfaction as covariates. I was interested in analyzing hazardous alcohol use as a moderator and found evidence that there is an interaction between alcohol use and perceived culture.

My question is, when I ran the regression the output did not show the ORs for hazardous alcohol use or perceived culture (scored from 0-4 based on how many items endorsed). Is there an option I am missing to get those estimates? (Code below)

Proc logistic data=c_final:
Class sex agecat race hazardousalc jobsat/param=ref; (I specified the reference groups in my code but omitted them for a reason on this forum)
Model departure = sex agecat race hazardousalc jobsat alcoholculture_score hazardousalc*alcoholculture_score/expb;
Run;

Re: Odds ratios not showing for individual variables in interaction term? Logistic regression

PaigeMiller — Thu, 21 Apr 2022 13:22:50 GMT

Try using the ODDSRATIO statement in PROC LOGISTIC

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_logistic_syntax26.htm

Re: Odds ratios not showing for individual variables in interaction term? Logistic regression

StatDave — Thu, 21 Apr 2022 13:37:11 GMT

As discussed in this note, odds ratios are not automatically provided for variables involved in interactions. It shows several ways to obtain odds ratios.

Re: Odds ratios not showing for individual variables in interaction term? Logistic regression

christinef — Thu, 21 Apr 2022 13:58:09 GMT

This worked with oddsratio hazardousalc/at (alcoholculture_score 0 1 2 3 4);

Thank you!