0
$\begingroup$

I have a zero-inflated negative binomial model. I have used incidence rate ratios and I'm trying to interpret the coefficients in relation to my predictors. Most of my predictors are continuous variables of census data -- ie: % of the population that is Hispanic; % of the population less than age 18, etc. I know that the IRR is normally interpreted as the rate ratio for a 1-unit increase in the independent variable, but what does this mean in terms of these continuous predictors -- does this mean the IRR is the estimated rate ratio for a 1% increase in % Hispanic. Is there a way I can scale this so it can be interpreted to be the estimated rate ratio for a 10% increase in the % Hispanic? Also, one of my IRR's is 20. Does that seem unusually high?

$\endgroup$
  • $\begingroup$ The last question cannot be answered without knowing more about your outcome and the predictors. A factor of 20 increase for 1 percentage point increase in X could make sense in some contexts. You might also find the second example useful: ats.ucla.edu/stat/stata/output/stata_nbreg_output.htm $\endgroup$ – Dimitriy V. Masterov Jul 9 '13 at 17:26
1
$\begingroup$

Does this mean the IRR is the estimated rate ratio for a 1% increase in % Hispanic?

Yes.

Is there a way I can scale this so it can be interpreted to be the estimated rate ratio for a 10% increase in the % Hispanic?

Divide the variable by 10 before you run your regression.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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