I am currently doing an interactions paper with cox models to assess time to criminal recidivism as an outcome, but am struggling with the interpretations.

My two variables that are part of the interaction are continuous count variables:

v1 - Ranges from 0-5

v2 - Ranges from 0-2

Here are basic findings from the cox model:

v1 - Coeff.=0.21, p<0.001, HR=1,234

v2 - Coeff.=0.409, p<0.001, HR=1.505

v1*v2 - Coeff.=-0.079, p=0.044, HR=0.92, CI - 0.875-0.976

How could I interpret this interaction term?

The best I've got right now is "The interaction of v1 and v2 was found to be significant (p-value=0.004) with a hazard ratio for the interaction term of 0.92 and a confidence interval from 0.875 to 0.976", but this doesn't say much at all...

Thanks for your help!


1 Answer 1


Since the coeficient is negative, while both main effects are positive, it means that each of the main effects is lessened by increases in the other one.

You can add up coefficients for different levels. When v1 = 5 and v2 = 2, the interaction is 10, and the total coef. is $5*0.21 + 2*.41 + 10*-0.079$. You can do this for every combination and make a table or graph of the results.

  • $\begingroup$ Thank you so much Peter! $\endgroup$
    – Raquel
    Jul 26, 2023 at 12:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.