1
$\begingroup$

This relates to the problem of calculating the probability of a coin's bias given an observed sequence of flips consisting of h head, and t tail flips. If the coin bias is unknown but known to be uniformly distributed over the interval [0,1], then the expectation for the bias (x) is given by summing the probability function over all possible values of the bias:

$ \Bbb E[x] = \int _0^1 x^h (1-x)^t dx $

I believe the result is:

$ \Bbb E[x] = \frac{h!t!}{(h+t+1)!} $

But I don't know how to derive that result. I'm trying to use integration of parts and not getting very far.

$\endgroup$
2
  • 2
    $\begingroup$ Can you use the beta function? $\endgroup$ Oct 8, 2016 at 20:23
  • 1
    $\begingroup$ If $t$ and $h$ are non-negative integers, successive integration by parts does the trick. The proposed solution needs a factorial on the denominator. $\endgroup$
    – Mark Viola
    Oct 8, 2016 at 20:34

2 Answers 2

2
$\begingroup$

HINT:

Integrate by parts with $u=x^h$ and $v=-\frac1{t+1}(1-x)^{t+1}$.

Repeat until the remaining Integral is $\int_0^1 (1-x)^{t+h}\,dx$

$\endgroup$
1
$\begingroup$

$$\int\limits_{0}^{1} x^{h} (1-x)^{t} \mathrm{d}x = \mathrm{B}(h+1, t+1)$$ where $\mathrm{B}(x,y)$ is the beta function, defined as $$\mathrm{B}(x,y) = \int\limits_{0}^{1} z^{x-1} (1-z)^{y-1} \mathrm{d}z$$ for $\Re x \gt 0$ and $\Re y \gt 0$. Since $h$ and $t$ represent coin flips, our expression in terms of the beta function is valid.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .