0
$\begingroup$

Let x ~ CN (0,a) and y ~ CN(0,b), with a, b positive constants. Let z = x + y. I need to find the distribution of the vector t=[x; z]. My basic understanding is that vector t is also complex Gaussian distributed and it has mean [0;0]. I do not know how to compute the covariance matrix of t. Can anybody help me? Thank you.

$\endgroup$
1
$\begingroup$

Because x and y are independent random variables, Var(z) = Var(x) + Var(y) = a + b;

One way to compute the covariance Cov(x,z) = Cov(z,x) is to simply apply the definition of covariance.

Cov(x,z) = E[xz*]-E[x]E[z*], and here we have E[x]=E[z*]=0.

Cov(x,z) = E[x(x+y)* ]=E[xx*+xy*]=E[xx*]+E[xy*]. Note that E[xy*]=E[x]E[y*]=0.

Cov(x,z) = E[xx*].

Observe that Var(x)= E[xx*]-E[x]E[x*] = 1, and E[x]=E[x*]=0 in this exercise.

Finally, Cov(x,y) = a.

Having computed the variances and co variances it is then straightforward to write the correlation matrix.

$\endgroup$
4
  • $\begingroup$ These are not the correct definitions for the Complex Normal distribution. See en.wikipedia.org/wiki/…. $\endgroup$ – whuber Jan 23 at 18:34
  • $\begingroup$ Isn't it so that I have used the definitions for complex normal variables, while links points to definitions for complex normal vectors? If I understand correctly, x, y and z are complex normal variables, and I compute variance and covariance for normal variables (not vectors). Am I wrong? Thanks $\endgroup$ – Rocco Jan 23 at 18:50
  • $\begingroup$ You are wrong because you neglect to take complex conjugates. $\endgroup$ – whuber Jan 23 at 18:51
  • 1
    $\begingroup$ OK, I see. I try to correct the answer then. $\endgroup$ – Rocco Jan 23 at 19:08

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.