# Tail bounds for Beta distribution

Say $X\sim\mathrm{Beta}(\alpha,\beta)$. Are there any "nice" closed form upper bounds for the tail probability $P(X\geq\epsilon)$, that are reasonably tight when $\beta$ is large? By "nice" I mean involving only elementary functions, and not for instance the incomplete beta function.

In my specific setting I have $\alpha=1/2$ and $\beta=d/2$ for some (typically large) integer $d$, and so one way of bounding the tail is to write $X=\frac{Y}{Y+Z}$ where $Y\sim\chi^2_1$ and $Z\sim\chi^2_d$, and using an upper bound on $Y$ and lower bound on $Z$ (e.g. from here: What are the sharpest known tail bounds for $\chi_k^2$ distributed variables?). This gives an answer, but it's somewhat messy (requiring the small $d$ case to be handled separately, etc). Is there a cleaner way?

Thanks.

• Do you intend $\epsilon$ to be close to $1$? Would it have any definite relationship to the parameters, such as exceeding the mean, median, or mode (to justify calling this a "tail probability")? – whuber Nov 12 '14 at 19:18
• In my specific case when $\beta$ is big and $\alpha$ is a constant, I'd want $\epsilon$ to be just slightly larger than the mean. A clearer way to state the question may be to say that for instance I want a bound like $X\leq f(\log(1/\delta))$ with probability at least $1-\delta$ for some monotonic function $f$ (e.g. $\log(1/\delta)/d$, or something of the sort). – martin Nov 12 '14 at 20:10

We write $I_{\epsilon}(A,B,\epsilon)$ the incomplete beta function. Thus $P(X>\epsilon) = 1-I_{\epsilon}(A,B,\epsilon)$

A simple bound is

$$\frac{2\sqrt{\epsilon}}{2\sqrt{\epsilon}+B\left(\frac{1}{2},\frac{d}{2}\right)}< I_\epsilon\left(\frac{1}{2},\frac{d}{2},\epsilon\right)<\frac{2 \sqrt{\epsilon}}{B\left(\frac{1}{2},\frac{d}{2}\right)}$$

A tighter lower bound (so upper bound for the tail) would be

$$\frac{2 \left(\sqrt{\epsilon} B\left(\frac{1}{2},\frac{d}{2}\right)+2 \epsilon\right)}{2 \sqrt{\epsilon} B\left(\frac{1}{2},\frac{d}{2}\right)+B\left(\frac{1}{2},\frac{d}{2}\right)^2+4 \epsilon}$$

Mathematica will spit them out for you with

1 - 1 / Normal[Series[1/(1 - BetaRegularized[\epsilon, 1/2, d/2]), {\epsilon, 0, n}]]

• How is $x$ related to the $\epsilon$ of the question? – whuber Nov 12 '14 at 19:16
• $x = \epsilon$ ... – Arthur B. Nov 12 '14 at 20:35
• $P(X \geq \epsilon) = 1-I_{\epsilon}\left(\frac{1}{2},\frac{d}{2},x\right)$ – Arthur B. Nov 12 '14 at 20:46
• Okay, so then what is $x$, since in your latest comment you seem to be distinguishing $x$ and $\epsilon$? – whuber Nov 12 '14 at 21:17
• No, it's always true – Arthur B. Nov 13 '14 at 12:06