Login  Register

RE: Using quantlib with MS VC++ version 7.1 for eigenvector calcs

Posted by Elliffe, Eoin on Apr 10, 2006; 8:47am
URL: http://quantlib.414.s1.nabble.com/Using-quantlib-with-MS-VC-version-7-1-for-eigenvector-calcs-tp4693p4694.html

What you say is certainly true, that is where V is the norm corr matrix.
V = W*diag(lambda)*W' even if the elements of W the eigenvector matrix
are negative.
I meant to say the eigenvalues are small and negative.
This is a problem....
 

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Ferdinando Ametrano
Sent: 07 April 2006 22:55
To: Elliffe, Eoin
Cc: [hidden email]
Subject: Re: [Quantlib-users] Using quantlib with MS VC++ version 7.1
for eigenvector calcs

On 4/7/06, Elliffe, Eoin <[hidden email]> wrote:
> I am using some algorithms from numerical recipes in C, namely tqli()
> and tridiag(), to calculate the eigenvectors/values of a corr matrix.
>
> I found that certain eigenvectors had elements with the correct
> magnitude but with opposite sign as those found using octave which
> uses hessenberg and schur decomposition.

why you consider this a problem?
Eigenvalues must be the same, eigenvector are defined with no regards to
their sign. If A is a matrix and k is one of its eigenvalue, there is an
eigenvector v of unit norm for which A v = k v holds.
The same equation holds also for  -v, which is of unit norm too. So v
and -v are completely equivalent.

btw 4v, 7v, 2.3v are eigenvector too, even if their norm != 1

> I decided to take a look at Quantlib to see if I could remedy this.
glad you trust QuantLib... but it's not going to remedy a non-issue

hope it helps

ciao -- Nando
_______________________________________________________________________

This e-mail may be privileged and/or confidential, and the sender does not waive any related rights and obligations.
Any distribution, use or copying of this e-mail or the information it contains by other than an intended recipient is unauthorized.
If you received this e-mail in error, please advise me (by return e-mail or otherwise) immediately.