Suppose that a factory produces coins whose bias θ follow a Beta distribution with parameters α=5 and β=10. If a random coin was chosen from the factory and tossed 100 times which resulted in 50 heads and 50 tails. What is the probability of coin being fair?

My approach:

Prior : Beta(5,10)

Posterior : Beta(50+5,50+10)

Now coin being fair is θ=0.5, But the probability of point in pdf is 0. I think the answer to 'probability of coin being fair' is 0. But still confused.

Am I missing something?


I think you've made a small type in your answer, the posterior is Beta(55, 60).

That aside, I think the question is (perhaps intentionally) vague. It requires you to do more than just maths, you need to interpret as well. You could argue that a coin is fair if $0.5 - \epsilon < p < 0.5 + \epsilon$, and then calculate

$\int _{0.5 - \epsilon}^{0.5+\epsilon}\beta(q ; 55, 60)dq$

for a range of $\epsilon$.

Or you could take the hardliner approach and say the probability of the coin being perfectly fair will always be zero if you use a Beta prior (as far as I can tell, you need a prior which has some sort of delta function like properties, i.e. some finite mass concentrated at a point, for there to ever be a finite probability of the coin being exactly fair)

| cite | improve this answer | |
  • $\begingroup$ Thanks. It was really helpful. PS I corrected the type $\endgroup$ – Shiva Shankar Aug 26 at 11:58

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.