2
$\begingroup$

I would like to compute $E[Xr / (Xr +1 - X)] $ where $X$ follows a Beta distribution $\operatorname{Beta}(\alpha, \beta)$ with $\alpha, \beta > 1$, $\alpha < \beta$ and $r \in (0,1)$. This is the same as computing the integral: $$\int_0^1 \frac{xr}{xr +1 -x} \frac{x^{\alpha-1} (1-x) ^ {\beta - 1}}{\operatorname{Beta}(\alpha,\beta)} \, dx$$ where $Beta$ is the beta function.

I'm looking for a closed formula that would be a "simple" function of $r$ and known functions, such as the $Beta$ or $\Gamma$ functions.

For instance, to solve $E[X] =\int_0^1 x \frac{x^{\alpha-1} (1-x) ^ {\beta - 1}}{\operatorname{Beta}(\alpha,\beta)} \, dx$, noticing that:

$\int_0^1 x^\alpha (1-x)^{\beta - 1} \, dx = \operatorname{Beta}(\alpha+1,\beta) = \operatorname{Beta}(\alpha,\beta) \frac{\alpha}{\alpha + \beta} $ (via the $\Gamma$ function)

gives the answer. But I can't find such a trick or a nice change of variable for $E[Xr / (Xr +1 - X)] $.

Thanks in advance

$\endgroup$
0
$\begingroup$

Have you tried using a computer algebra system to help see if a 'closed-form' might exist? For example, given $X \sim \text{Beta}(a,b)$ with pdf $f(x)$:

enter image description here

... you seek:

enter image description here

This is a 'closed-form' solution, but perhaps not the simple solution you might have hoped for.

Notes:

  1. The Expect function used above is from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors.

  2. For more detail on the Hypergeometric2F1Regularized function, see, for instance: http://reference.wolfram.com/language/ref/Hypergeometric2F1Regularized.html

$\endgroup$
  • $\begingroup$ It seems to indicate there's no easy closed form, doesn't it? I should have been more specific: basically I want, given r, alpha and beta (which can take different values) the expectation in the most efficient way with a computer. So I'll check if that function you mention is implemented in Java package for instance and how it performs. Thanks for your answer! $\endgroup$ – Michael May 20 '16 at 13:50

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.