BDA3 Chapter 2 Exercise 2

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

\(\DeclareMathOperator{\dbinomial}{binomial} \DeclareMathOperator{\dbern}{Bernoulli} \DeclareMathOperator{\dbeta}{beta}\)

We are given the following information about the two coins.

\[ \begin{align} p(C_1) &= 0.5 & p(H \mid C_1) &= \dbern(H \mid 0.6) \\ p(C_2) &= 0.5 & p(H \mid C_2) &= \dbern(H \mid 0.4) \end{align} \]

The posterior probability of each coin given two tails is:

\[ \begin{align} p(C_1 \mid TT ) &\propto p(TT \mid C_1) \cdot p(C_1) \\ &= \left(\frac{2}{5}\right)^2 \frac{1}{2} \\ &= \frac{2}{25} \end{align} \] \[ \begin{align} p(C_2 \mid TT ) &\propto p(TT \mid C_2) \cdot p(C_2) \\ &= \left(\frac{3}{5}\right)^2 \frac{1}{2} \\ &= \frac{9}{50} \end{align} \]

Both of the previous probabilities are normalised by the same constant. Since \(p(C_1 \mid TT) + p(C_2 \mid TT) = 1\), the normalising constant is \(\frac{2}{25} + \frac{9}{50} = \frac{13}{50}\). Thus

\[ p(C_1 \mid TT) = \frac{4}{13} \qquad \text{and} \qquad p(C_2 \mid TT) = \frac{9}{13}. \]

Let \(y\) be the number of additional spins until the next head. Conditional on a coin, \(y\) is geometrically distributed. So the expected number of spins before the next head is:

\[ \begin{align} \mathbb E(y \mid TT) &= \frac{4}{13}\mathbb E(y \mid C_1) + \frac{9}{13}\mathbb E(y \mid C_1) \\ &= \frac{4}{13}\frac{5}{3} + \frac{9}{13}\frac{5}{2} \\ &= \frac{20}{39} + \frac{45}{26} \\ &= \frac{175}{78}, \end{align} \]

which is 2.24359.