2
$\begingroup$

Given a distribution $P_X$ on $\mathbb{R}$, when does there exist a coupling (i.e. joint distribution) $P_{X^n}$ of $X_1,...,X_n$, each distributed according to $P_X$, such that $\sum X_i^2 = n$ almost surely?

The motivation comes from the following: let $P_{X^n}$ be an arbitrary distribution on $\sqrt{n}S^{n-1}$ that is permutation invariant, i.e. we can assume all marginal distributions $X_i$ have the same probability law. Then how can we characterize the marginal distribution of $P_{X^n}$? E.g. Clearly $\mathbb{E}[X^2] = 1$, and $X^2 \leq n$ a.s., but given a random variable satisfying these properties, can I couple $n$ copies of them onto the sphere?

$\endgroup$
3
  • $\begingroup$ The class of such distributions will probably be fairly large. For example, for each i.i.d. sequence $X_i$ with $\mathbb{P}(X_1 = 0) = 0$ the distribution of $\sqrt{n}X_i\left(X_1^2 + \ldots + X_n^2\right)^{-\tfrac{1}{2}}$ satisfies the condition. Why are you only interested in those distributions that have the same marginals for all coordinates? This property is certainly less restrictive than the permutation invariance. $\endgroup$
    – Dominik
    Aug 26, 2015 at 22:34
  • $\begingroup$ You shouldn't expect to be able to do this for atomic distributions. If you had a distribution supported on values whose squares never sum to $n$, you would have a counterexample. The same would be true if the distribution is `almost atomic' (e.g. 99.99% of the mass is within $10^{-100}$ of a collection of values with the above property. $\endgroup$ Aug 26, 2015 at 22:54
  • $\begingroup$ These are good points. Some interesting examples are found by taking a dirac delta $\delta_{x^n}$, and averaging it over all permutation. Convex combinations of such averages preserve $\|X^n\|^2 = n$. And indeed we can't recover a distribution of this form by Dominik's methods. I'm interested in marginals since I'm interested in which distributions $P_Y$ can be induced through a transform $P_{Y|X}$ of the marginal. Specifically, I'm maximizing some function of $P_Y$ over such distributions. But I thought the question of which distributions could be coupled is interesting alone. $\endgroup$
    – AustinC
    Aug 26, 2015 at 23:39

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.