We've had the following question discussed today but without any result:

Let $X_1,\dots,X_d$ be random variables, iid and $X_n\sim N(\mu_n,1)$.

How can we describe the distribution of $\|P_VX\|^2$ with $X=(X_1,\dots,X_d)$, $V\subset \mathbb R^d$ and an orthogonal projection $P$?


Hint: use the fact that $$ P_V(I - P_V) = 0 $$ and the Cochran theorem.

  • $\begingroup$ That is the standard one, at least. This is a very classical approach presented in many books of statistics. $\endgroup$ – mookid Oct 29 '14 at 23:07
  • $\begingroup$ My problem is to find the $q_i$'s, how do you get them? $\endgroup$ – Ele Oct 29 '14 at 23:16
  • $\begingroup$ The version of the theorem in the link is not the most clear I have seen, I'm sorry for that. I'll give a working version here: if $X\sim N(\mu,\Sigma)$ and if you can find orthogonal projectors $p_i$ whose sum is $id$ and such as $i\neq j\implies p_i\circ p_j = 0$, then the variables $p_i X$ are independant, with marginals $N(p_i\mu, p_i \Sigma p_i)$ $\endgroup$ – mookid Oct 29 '14 at 23:27

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.