6
$\begingroup$

Often, one is often interested in theorems/inequalities of the following kind:

Let $X$ be a random variable then the probability that $X$ is close to typically $\mu$ (or larger than some constant) is small. In other words the theorem shows that $X$ decays rapidly in some sense. A typical example is Chebyshev's inequality, which states essentially that $X = \mathbf E[X] + O(\sqrt{\mathbf{Var}[X]})$.

Chapter 1 in Additive combinatorics by Tao and Vu provides lots of interesting results of this kind. (My own interest comes from combinatorics.) My question concerns theorems of the opposite kind:

What are some good theorems that can be used to show that $X$ decays slowly?

For instance, I would be interested in theorems of the sort $\mathrm{Pr}( |X-\mu|> a) > b$ (or without the absolute value), where $a$ and $b$ depend on the moments of $X$. (I am particularly interested in discrete distributions.)

One extremely simple example is: $$ \mathrm{Pr}\big( |X - \mathbf E[X]| > \mathbf{Var}(X)^{1/2} \big) > 0.$$

But surely there must be many others?

$\endgroup$
2
$\begingroup$

The Paley-Zygmund inequality says that if $0 < \lambda < 1$, $X \ge 0$, and $E(X^2) < \infty$, then $$ \Pr(X \ge \lambda E(X)) \ge (1-\lambda)^2 \frac{(EX)^2}{E(X^2)} .$$ It is in the book by JP Kahane "Some Random Series of Functions," at the end of Section 6 of Chapter 1.

I have some papers that gives a lower bound on tail probabilities, typically on sums of independent random variables. The first paper is based upon the inequality of Paley and Zygmund. http://faculty.missouri.edu/~stephen/preprints/tail.html http://faculty.missouri.edu/~stephen/preprints/compare.html http://faculty.missouri.edu/~stephen/preprints/weibulls.html http://faculty.missouri.edu/~stephen/preprints/tailproc.html http://faculty.missouri.edu/~stephen/preprints/disttail.html I know a lot of other people have done this type of thing as well - look at the references in the last paper. In particular, I was really impressed by the paper by Latala, (1997) Estimation of moments of sums of independent random variables, Ann. Probab., 25, 1502-1513.

$\endgroup$
2
  • $\begingroup$ The links are not available: "Authorization Required" $\endgroup$ Apr 22 '17 at 16:20
  • $\begingroup$ @LeGrandDODOM Should be fixed now. $\endgroup$ Apr 22 '17 at 20:11

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.