There is the following task: Suppose we toss a coin $ N = 10$ times and observe $m = 9$ heads. Let the null hypothesis be that the coin is fair,and the alternative be that the coin can have any bias, so $p() = U(0; 1)$. Derive the Bayes factor $BF_{1;0}$ in favor of the biased coin hypothesis. What if $N = 100$ and $m = 90$?

What I did so far:

Prior likelihood if coin is biased or unbiased is $0.5$ For the unbiased coin I used binomial distribution:

$p^m (1-p)^{(N-m)}$ = $(0.5)^{10}$

But how to count for the biased coin? With what probabilities should I count?


1 Answer 1


I think that the wikipedia article about Bayes factor presents an example at the end that is exactly the same as yours (https://en.wikipedia.org/wiki/Bayes_factor).

This is how I understand it and would apply it to your problem: if $H_0$ is the unbiased coin hypothesis then $P(N=10\ m=9\ | H_0) = \binom{10}{9}\frac{1}{2}^{9}\frac{1}{2}^{10-9}$

If $H_1$ is the hypothesis that the coin has a bias $p \in [0,1]$ and that $p$ can have any value in $[0,1]$ with uniform probability ($U(0;1)$) then $P(N=10\ m=9\ | H_1) = \int_0^1 \binom{10}{9} p^9(1-p)^{10-9}dp$

Then the Bayes factor is given by the ration of the two probabilities.

Note: by integration by parts $\int_0^1 p^a(1-p)^{b}dp = \frac{b}{a+1}\int_0^1 p^{a+1}(1-p)^{b-1}dp = ... = \frac{b!}{(a+1)...(a+b)}\int_0^1p^{a+b}dp = \frac{1}{(a+b+1)\binom{a+b}{b}}$


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.