Let $X$ be a random vector in $\mathbb{R}^n$ whose entries are joint Gaussian with zero mean and covariance matrix $K.$ Is there a closed form expression for $\mathbb{E}||X||_2,$ as there is for the absolute deviation of a standard Gaussian in a 1-dimensional space?

  • $\begingroup$ what is $\|X\|_2$? $\endgroup$ – Conrado Costa Aug 24 '15 at 19:34
  • $\begingroup$ $\sqrt{\sum_i X_i^2}$ $\endgroup$ – Hedonist Aug 24 '15 at 20:21
  • $\begingroup$ You can use the so called Khinchine-Kahane inequality to show that there are universal constants $c_1,c_2>0$ so that $c_1 \le E\|X\|_2 / \sqrt{E\|X\|_2^2} \le c_2$. $\endgroup$ – Stephen Montgomery-Smith Aug 24 '15 at 21:52
  • $\begingroup$ I believe the constants $c_1,c_2$ will depend on the dimension of the vector. I would like to have dimension free estimates if possible. Thanks. $\endgroup$ – Hedonist Aug 25 '15 at 12:16
  • $\begingroup$ No, the constants won't depend upon the dimension of the vector. $\endgroup$ – Stephen Montgomery-Smith Aug 25 '15 at 13:44

If you can settle with a diagonal covariance matrix, then please check "Multidimensional Gaussian Distributions" by Kenneth S. Miller (1964 edition, chapter 2, section 2, RAYLEIGH DISTRIBUTIONS). Otherwise you need to deal with a lot more complicated equations. This reference could be a good start :

"Properties of Generalized Rayleigh Distributions"
L. E. Blumenson and K. S. Miller
The Annals of Mathematical Statistics
Vol. 34, No. 3 (Sep., 1963), pp. 903-910

You can find a copy of this paper at JSTOR (free sign up!).

  • $\begingroup$ Thanks Ali. My covariance matrices are not diagonal. I would also like to point out a previous post that solves the problem for identity matrices (your answer subsumes it though): math.stackexchange.com/questions/827826/… $\endgroup$ – Hedonist Aug 25 '15 at 17:56
  • $\begingroup$ Unfortunately, where the author of the book talks about "diagonal covariant matrix" (prior to Theorem 1 of Ch. 2 sec 2), he then proceeds to write a scalar multiple of identity. $\endgroup$ – user357151 Dec 19 '17 at 21:42

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.