MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top


I wish to calculate (or upper bound) expectations of the form $E[\langle x,y \rangle^2]$, where $x$ and $y$ are i.i.d standard gaussian vectors of length n. Are there any exponential type upper bounds for the same?

share|cite|improve this question
WHat is an "exponential-type upper bound"? – Igor Rivin Nov 25 '11 at 21:51
an upper bound that exponentially decays to 0 – user19530 Nov 25 '11 at 22:36
I think I am misunderstanding the question along the same lines as Igor Rivin's answer below... another way of phrasing his answer is: let $W_i = x_iy_i$, then since each $W_i$ has mean zero you are looking at the variance of a sum of $n$ i.i.d. copies of $W_1$, and $W_1$ has variance $1$, so the whole sum has variance $n$. – Yemon Choi Nov 26 '11 at 0:05

I must be misunderstanding the question, but $<x, y>^2$ is a sum of the terms of the form $x_i x_j y_i y_j.$ The expectation of this term vanishes, unless $i=j,$ in which case it (the . expectation) is the square of the expectation of the square of the standard Gaussian. The square of the standard gaussian is the chi-square distribution with one degree of freedom, whose mean is $1,$ so the whole thing should be $n$.

share|cite|improve this answer
Perhaps I didnt pose the question correctly. What I am looking for is something along these lines: let $\theta = <x,y>$, where x and y are as defined earlier. so $\theta$ is a sum of bessel functions, whose mean is 0. I am now looking to upper bound the expected value of $\theta^2$ . – user19530 Nov 25 '11 at 22:35
Perhaps you can explain Igor's misunderstanding, since I have it too. – Brendan McKay Nov 26 '11 at 14:59

Your Answer


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.