Consider an offset term in a Poisson regression:

$$\log \mu_x = \log t_x+ \beta_0 + \beta_{1} x$$

To interpret $\beta_0$, would you need to consider $(\beta_0+ \log t_x)$? Because what if $t_x \neq 1$? Is the interpretation of $\beta_0$ the mean number of events when $t_x = 1$ and $x=0$? Suppose $t_1 = 2, t_2= 3$ and $t_3 = 4$. Then there is no $x$ such that $t_x = 1$. Also $x \neq 0$.

Also is the interpretation of $\beta_1$ the following: The mean number of events comparing $x+1$ and $x$ for a fixed $t_x$?


1 Answer 1


An alternative formulation of your model is $$ \log\left(\frac{\mu_x}{t_x}\right) = \beta_0 + \beta_1 x $$ Then, we can see that $$ \beta_0 = \log\left(\frac{\mu_0}{t_0}\right) $$ and $$ \beta_1 = \log\left(\cfrac{\mu_{x+1}}{t_{x+1}}\right) - \log\left(\cfrac{\mu_{x}}{t_{x}}\right) = \log \left( \cfrac{\,\,\,\,\cfrac{\mu_{x+1}}{t_{x+1}}\,\,\,\,}{\cfrac{\mu_{x}}{t_{x}}} \right) $$ In words,

  • $\beta_0$ is the log rate at $x=0$;
  • $\beta_1$ is a log rate ratio ($x+1$ versus $x$)
  • 2
    $\begingroup$ +1, but I wonder if your equation for $\beta_0$ could be clarified... $\endgroup$ Jun 1, 2014 at 18:19

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.