Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $X \sim \operatorname{Binom}[(n,p)]$ and $Y \sim \operatorname{Poisson}[f(X)]$, where f is a convex function. Are there any good tail bounds for $Y$? For instance, are there any Chernoff-style bounds for $Y$?

share|improve this question
    
The article "The moment bound is tighter than Chernoff's bound for positive tail probabilities", referenced at this post may be of interest. –  Sasha Nov 7 '11 at 21:27

1 Answer 1

The Cheroff's bound, for positive random variable, and arbitrary $\theta$ such that the moment generating function $\mathcal{M}_Y(\theta)$ exists: $$ \mathbb{P}(Y \ge t) \le \mathrm{e}^{-\theta t} \mathcal{M}_Y(\theta) = \mathbb{E} \left( \mathrm{e}^{-\theta t + f(X) \left( \mathrm{e}^\theta - 1 \right))} \right) $$

If $g_\theta(X) = \mathrm{e}^{-\theta t + f(X) \left( \mathrm{e}^\theta - 1 \right))}$ happens to be concave, then by Jensen's inequality we would have: $$ \mathbb{P}(Y \ge t) \le \mathrm{e}^{-\theta t + f( \mathbb{E} \left(X\right)) \left( \mathrm{e}^\theta - 1 \right))} = \mathrm{e}^{-\theta t + f( n p ) \left( \mathrm{e}^\theta - 1 \right))} $$

The inequality above would be true for any such $\theta$ that $g_\theta^{\prime\prime}(X) \le 0$ for all $X \ge 0$: $$ 0 \ge g_\theta^{\prime\prime}(x) = g(x) \left\{ \left( f^\prime(x) ( \mathrm{e}^{\theta} - 1) \right)^2 + f^{\prime\prime}(x) ( \mathrm{e}^{\theta} - 1) \right\} $$

Alternatively one could look at the moment bound:

$$ \mathbb{P}(Y > t) \le \frac{1}{t} \mathbb{E}(Y) = \frac{1}{t} \mathbb{E}(f(X)) $$

Again, if $f(x)$ is concave, by Jensen's inequality we would have $$ \mathbb{P}(Y > t) \le \frac{f(n p)}{t} $$

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.