I'm referring to page 22 of this white paper.

On page 22, it says the following: given that

  • $s_i \sim \text{Exp}(\theta), i = 1,..,c$

  • $\theta \sim\text{Gamma}(k, \Theta)$,

Then the posterior distribution for $\theta$ given the data is

$ \theta | s \sim \text{Gamma}(k + c, \frac{\Theta}{1 + \Theta cs}) $

where $s$ is the mean of the data, $s = \frac{\sum_{i=1}^{c}s_i}{n}$

But this is inconsistent with my understanding of the conjugate Gamma-Exponential system, where after updating the posterior distribution is as follows:

$ \theta | s \sim \text{Gamma}(k + c, k + cs)$

Am I missing something super obvious here? Thank you!


Two things.

  1. I think you have a typo: you're probably expecting the posterior to be $\text{Gamma}(k+c, \Theta + cs)$.

  2. Note that the author of the paper is using the alternative parametrization of the exponential distribution, which has pdf $f(x) = \frac{1}{\theta}e^{-x/\theta}$. If you're looking at a derivation that uses the standard characterization, you might have some flipped fractions. In particular, try looking at the reciprocal of $\frac{\Theta}{1 + \Theta cs}$.


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.