# Calculation of an expectation for the 'part' of a vector

Let $x$ be vector in $R^n$. Let $\pi(⋅)$ be a permutation on the set $\{1,\ldots,n\}$ with a uniform distribution. Let $|m|\leq n, m \in Z$.

Calculate $$E\left|\sum_{i=1}^mx_{\pi(i)}\right|^q, \quad q\geq 2.$$

Thank you.

-
There are $n!$ permutations, and $2^n$ subsets, of the set in question. The latter number is much smaller than the former, and still smaller is $\binom n m$, the number of subsets of size $m$. And you're really just choosing a random subset of size $m$ and working with that. It doesn't matter in which order you place the element within that subset or within its complement. But so far I don't really know if that helps. – Michael Hardy Jan 31 '13 at 3:47
Without the absolute value and the $q$th power, this would be really easy. – Michael Hardy Jan 31 '13 at 3:59
Thank you. Yes, I've been analusing this, but did not get any result. Is it possible at least to get a good bound from above? – Alex Jan 31 '13 at 4:03

If $n\geqslant2$ and $q=2$, the result is $$\frac{m(n-m)}{n(n-1)}\sum_{k=1}^nx_k^2+\frac{m(m-1)}{n(n-1)}\left(\sum_{k=1}^nx_k\right)^2.$$ A similar (but more tedious) expansion is possible for $q=4$ and more generally for every even integer $q$.
Since I do not understand the condition $|m|\leqslant n$, this assumes that $0\leqslant m\leqslant n$. – Did Jan 31 '13 at 12:12
@Did: Thank you. Is there any possibility to approimate the $q$-th moment $E|\sum_{i=1}^mx_{\pi(i)}|^q$ using bernoulli (or maybe some other) distribution? Thank you. – Alex Feb 20 '13 at 16:46