# Euclidian norm of Gaussian vectors

Let $X \sim \mathcal{N}(0, \Sigma)$ be a Gaussian vector in dimension $N$. I am interested by the probability density of the random variable $\lVert X \lVert_2$.

If $\Sigma = {I}_N$, we recognize the $\chi$-law. We especially know that the probability density is given by $$p(x) \propto x^{{N} -1} \mathrm{e}^{-\frac{x^2}{2}} 1_{x\geq 0} .$$

In the general case, we can decompose the matrix $\Sigma = P^t D P$ with $P$ orthogonal and $D=D(\lambda_1,\cdots,\lambda_N)$ diagonal and $\lVert X \lVert_2 \sim \lVert \mathcal{N} (0,D)\lVert_2$. What can we say about the probability density of $\lVert X \lVert_2$ in this general case?

After you diagonalize the covariance matrix, you have $||X||^2=\sum_{k=1}^n X_i^2$, right? And $X_1^2,\ldots,X_n^2$ are independent $\chi(1)$-distributed r.v. (up to a scaling with $\frac{1}{\sqrt{\lambda_i}})$, i.e., they are gamma-distributed, but with different scale parameter. So you can apply the results of this paper:
• Effectively, the paper of Moschopoulos (and actually the other one too) gives a formula for the density of $\lVert X \lVert_2^2$ from what we can deduce another one for $\lVert X \lVert_2$. It is expressed as an infinite sum. Thanks a lot for these references. Jan 30, 2014 at 15:20