# How to interpret logarithmically transformed coefficients in negative binomial regression?

How can I interpret log-transformed independent variables in terms of percent change in a negative binomial regression?

• Using what link function? – Glen_b Sep 30 '14 at 23:24
• @Glen_b: Let's say it's the log link. – Brash Equilibrium May 11 '15 at 20:15
• @Brash in that case, in much the same way you'd interpret it in a log-log regression (with the additional twist that it's percentage change in counts). I guess I'll post an answer. – Glen_b May 12 '15 at 1:53

ii) if some other link is used, some other interpretation would be needed -- but again, the interpretation is again similar to in the regression case; for example, an identity link would yield a similar coefficient-interpretation as for an ordinary regression where $x$ is logged but $y$ is not.