Skip to main content
2 of 3
deleted 201 characters in body
mermeladeK
  • 345
  • 2
  • 14

To add some insights into my discussion. I found out some stuff that might be interesting according to matrix perturbation theory.

So instead of asking as in the 1st question: given prescribed eigenvalues what is the minimum $c_{max}$? I tried to look into the problem of what is the possible eigenvalue densities given I have a prescribed $c_{max}$.

It goes like this... I compare my correlation matrix with the identity matrix. The diagonal elements of both matrices are 1's, therefore the perturbation is in the off-diagonal elements. The Weyl-Lidskii theorem states that:

$|\lambda_i-1|\leq ||\mathbf{C}-\mathbf{I}||_2$

where I write $|\lambda_i-1|$ because the eigenvalues of the identity matrix $(\mathbf{I})$ are $1$. However getting further than here does not seem trivial to me.

mermeladeK
  • 345
  • 2
  • 14