Let's say that I want to check the associations of mild-exercise to hypertension. I have a set of variables (apart from the IV Of interest and DV):

a. sex

b. race

c. smoking status

In my experience, it's better to look up for interaction first before drawing conclusion of confounding. Here are my steps:

a. look up for cOR for mild-exercise to hypertension (STATA command: logit hypertension i.mild-exercise)

b. look up for interaction between sex and mild-exercise to hypertesion (STATA command: logit hypertension i.mild-exercise##i.sex)

When I am looking for interaction between sex and hypertension, the OR drastically changes for male (>10%), but not with female (

Similar questions and discussions