I'm not sure if this is how you analyze confounding with possible interaction (effect modification). The predictor is opioid usage while the outcome variable is prevalence of prior injection drug usage. I am creating my models through logistic regression
Only confounder in model was SES in model, possible effect modifier was ethnicity. This was the SAS code I used looking at confounding and possible effect modification.
proc logistic data=work.ex4;
class opioid_use SES eth_2 (ref= '3. Caucasian')/ param=ref;
model ever_inject = opioid_use SES eth_2;
run;
Based on multivariate logistic model, I don't have effect modification only confounding
