Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

suppose x is a gaussian vector (0 mean, known covariance $\Sigma$). Suppose A is a gaussian matrix (the kinds used in compressed sensing). Will Ax be a gaussian vector with mean 0 and a random covariance $A \Sigma A^T$? or will the distribution be something entirely different?

share|improve this question
    
By symmetry $Ax/\|Ax\|$ will be uniformly distributed on the unit sphere. –  p.s. Nov 26 '11 at 19:16

1 Answer 1

up vote 4 down vote accepted

The question as you've asked it is a little bit confused. You declared that $A$ is a random matrix, and then asked if the product $Ax$ had a normal distribution with covariance matrix $A\Sigma A^T$. But that would itself be a random matrix - and that doesn't make any sense!

To make sense of your question, let's reduce it to the simplest case where $A$ is 1 x 1 and $x$ is a length 1 vector, and to keep it really straightforward suppose that the distributions of $A$ and $x$ are both standard normal.

Then the product $y=Ax$ follows the product normal distribution which is a scaled version of the modified Bessel function of the second kind.

So, no, in the more general case where $A$ is $m \times n$ and $x$ is in ${\bf R}^n$, $y=Ax$ will not be normally distributed.

share|improve this answer
    
+1. Learn something new today. –  Tim Nov 26 '11 at 19:31
    
+1 for the reduction to the simplest case. –  Did Nov 26 '11 at 19:48

Your Answer

 
discard

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.