Bill Harris wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> "John Randall" <[EMAIL PROTECTED]> writes:
>
>> You are using LAPACK correctly. The problems are:
>
> Well, almost. :-)
>
>> v=:{."1 (2{::dgeev_jlapack_ m2) NB. principal eigenvector
>> v
>> 0.414299 0.731056 0.106322 0.53161
>
> I've still got to learn these doubly inflected symbols (map and fetch),
> but I confused myself by thinking I was looking for row, not column,
> vectors.
>
> Had I seen I was only off by a constant factor, I wouldn't have
> worried, and I might even have discovered his normalization.
>
> I'm curious, though: is there a reason you knew to take the right
> eigenvector instead of the left except that it was clearly the one that
> was a multiple of Coyle's? If I try to apply this elsewhere, I won't
> have the fortune of having the answer.
>
"Eigenvector" almost always means "right eigenvector". I guess LAPACK's
method discovers the left ones as well.
>> (a) The original matrix is wrong.
>
> What do you mean by this? I think I missed something.
>
My apologies: I typed it in wrong.
>> (b) Coyle's data is weirdly normalized.
>
> Based on what he's using it for, I can see him normalizing the sum of
> the components rather than the magnitude of the vector. Non-technical
> clients will likely understand each element as what fraction of their
> total value applies to each component.
>
I realized this after I sent the message.
Best wishes,
John
> Thanks to everyone who replied; this was quick and useful help!
>
> Bill
> - --
> Bill Harris http://facilitatedsystems.com/weblog/
> Facilitated Systems Everett, WA 98208 USA
> http://facilitatedsystems.com/ phone: +1 425 337-5541
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (MingW32)
> Comment: For more information, see http://www.gnupg.org
>
> iD8DBQFEBjFY3J3HaQTDvd8RAlxxAJ9w+10KiyLE4P1U0nb2AmGJT+JSpQCfREBA
> QktfQcee0kdlrgA/hm9kZQQ=
> =5bik
> -----END PGP SIGNATURE-----
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
