Original problem: A point X is randomly chosen from the interval (0,1). Suppose X=x is observed. Then a coin with P(Heads) = x is tossed independently n times. Let Y be the number of heads in n tosses. Find the unconditional distribution of Y.

I've setup the problem below using total probability and a uniform pdf of x (a confirmation others agree my formulation is correct would be nice), but not sure how to integrate this:



To approach this, I would apply the binomial theorem, which holds for non-negative integer $c$:

$$ (a+b)^c = \sum_{i=0}^c {c\choose i} a^ib^{c-i} $$

When you apply this identity to $(1-x)^{n-y}$, the integrand becomes a standard polynomial in $x$:

\begin{align*} Pr(Y=y) &= {n\choose y}\int_0^1 \sum_{i=0}^{n-y} {n-y\choose i}(-1)^ix^{i+y}~dx \\ &= {n\choose y}\sum_{i=0}^{n-y}{n-y\choose i}(-1)^i\frac{x^{i+y+1}}{i+y+1}\bigg|_0^1 \\ &= {n\choose y}\sum_{i=0}^{n-y}\frac{{n-y\choose i}(-1)^i}{i+y+1} \end{align*}

  • $\begingroup$ wow nice....much appreciated -- i learned something new here! $\endgroup$ – JPJ Jun 12 '16 at 18:01
  • 1
    $\begingroup$ That also simplifies to $ \dfrac{1}{n+1}$ $\endgroup$ – Henry Jun 12 '16 at 23:59

I just realized that the integral is an archetype of a beta distribution and thus can be manipulated as such...beta distribution

$\int_{0}^{1}\begin{pmatrix}n\\y\end{pmatrix}x^y(1-x)^{n-y}\,dx$ ~ $\frac {\Gamma(\alpha+\beta)}{\Gamma(\alpha)\Gamma(\beta)}x^{\alpha-1}(1-x)^{\beta-1} $$\Rightarrow$ which integrates to 1 over the support (0,1).

Thus, if $\alpha-1=y$ and $\beta-1=n-y$ $\implies$ $\begin{pmatrix}n\\y\end{pmatrix}$$\frac{\Gamma(y+1)\Gamma(n-y+1)}{\Gamma(n+2)} $

I'm sloppy so hope I didn't make an algebra mistake. It should (I hope!) come out to the same thing @josliber wrote, and he still gets the credit because he managed more creativity IMO.

  • 4
    $\begingroup$ $\displaystyle {n \choose y}\frac{\Gamma(y+1)\Gamma(n-y+1)}{\Gamma(n+2)}$ can be simplified to $\dfrac{1}{n+1}$ for integers $\endgroup$ – Henry Jun 12 '16 at 22:14

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.