Given the following generative model

$$T | \mathbf{x_i} \sim \operatorname{Exp}(\lambda_i)$$ $$\ln(\lambda_i) = \beta_0 + \beta_1 x_1 + \beta_2 x_2 + \dots$$

That is: I have observations of covariate/response pairs $(\mathbf{x}_i, t_i)$, such that $t_i$ are measurements of inter-arrival time of events, and is assumed to be exponentially distributed. I would like to fit a GLM such that the response distribution is exponential and using a log link. I note that you can do it in a roundabout way using a gamma family: Fitting exponential (regression) model by MLE?

What I am wondering is whether it is possible to transform this into a Poisson regression model like the following:

$$Y | \mathbf{x_i} \sim \operatorname{Pois}(\lambda_i)$$ $$\ln(\lambda_i) = \beta_0 + \beta_1 x_1 + \dots + \beta_n x_n + \ln(t_i)$$

that is, include the time to observe one event as an offset term in the regression equation, which factors in the exposure time: Where does the offset go in Poisson/negative binomial regression?


Your Answer

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

Browse other questions tagged or ask your own question.