BDA3 Chapter 2 Exercise 18

Here’s my solution to exercise 18, chapter 2, of Gelman’s Bayesian Data Analysis (BDA), 3rd edition. There are solutions to some of the exercises on the book’s webpage.

Suppose we have n observations from a Poisson likelihood, yi∣θ∼Poisson(xiθ), with rate θ and exposure xi. We show that with a gamma prior, θ∼gamma(α,β), the posterior also has a gamma distribution.

As shown in the book, the likelihood and prior are

p(y∣θ)∝θ∑n1yie−θ∑nixip(θ)∝θα−1e−βθ.

Thus the posterior is

p(θ∣y)∝θ∑n1yie−θ∑nixi⋅θα−1e−βθ=θα−1+∑n1yie−θ(β+∑n1xi),

which shows that p(θ∣y)∼gamma(α+∑n1yi,β+∑n1xi).