Re: [Bug-gsl] [PATCH] a one-pass run algorithm for sigma (histograms)

[Johannes Buchner]

Hi Brian,

I looked into the algorithm in the mentioned reference.
However, the formula 1.6b, for n values at x_k, boils down to 
 \sum_(1+k)^(n+k){1/l}. This can be calculated if there is a discrete
weight (n values at x_k). I was unable to find a shorthand formula for
this sum (so it can be generalized from n to a continuous value).
Thus I conclude that it is not possible to use the mentioned algorithm
for continuous histogram weights.

However, the section mentions that there are other algorithms
available. If you could point me to one, I may be able to further look
into it.

Regards,
Johannes

On Tue, 10 Nov 2009 17:34:33 +0100
Brian Gough <address@hidden> wrote:

> At Mon, 9 Nov 2009 17:16:56 -0600,
> Johannes Buchner wrote:
> > 
> > >From 777fab8add4cc83df9c5148b84cad0fd28160e66 Mon Sep 17 00:00:00
> > >2001
> > From: Johannes Buchner <address@hidden>
> > Date: Sun, 8 Nov 2009 23:41:15 +1300
> > Subject: [PATCH] a one-pass run algorithm for sigma
> 
> Thanks for the email.  I think the algorithm referenced is different
> from that, to avoid the cancellation error in subtracting the sums of
> squares.
> 
> -- 
> Brian Gough
> 
> GNU Scientific Library -
> http://www.gnu.org/software/gsl/


-- 
Emails können geändert, gefälscht und eingesehen werden. Signiere oder
verschüssele deine Mails mit GPG.
http://web.student.tuwien.ac.at/~e0625457/pgp.html

pgpakOKx5c312.pgp
Description: PGP signature