Re: Histogram with negative bin content

From: Arthur E. Snyder <>
Date: Tue, 5 Sep 2006 14:05:50 -0700 (PDT)

Sorry, I've not sure what TSpectrum does. The only peaking find I've done was something home brewed (scan starting points and fit for Gaussian) I used for a toy mc illustration of the dangers of bump hunting. It would not have been bothered by negative bins.

It seems to me that setting bins to zero may not do much harm in your case. If you were fitting for a level or something like that it would bias your result, but in looking for a peak it seems like the worse that could happen is you might miss a few marginal ones that were dependent on a downward fluctuation to look like a peak.

-Art S.

A.E. Snyder, Group EC                        \!c*p?/
SLAC Mail Stop #95                          ((.   .))
Box 4349                                        |
Stanford, Ca, USA, 94309                      '\|/`                 o
phone:650-926-2701                              _          BaBar
FAX:650-926-2657                          Collaboration

On Tue, 5 Sep 2006, Marco Calviani wrote:

> Hi Arthur,
> thank you for your advice on negative bin content.
> The fact is that i've encountered a problem using the TSpectrum class
> on a histogram with some negative bin content (but just because it was
> a difference between a "real" physical histogram and a TSpectrum
> generated background histogram). The peak recognition is correct on
> the original histogram but it stop after "some" peaks in the
> original-background histogram (unfortunately also correcting for
> negative bins does not solve the problem). You can see in the
> attachment the original (c1.jpg, ok) and the produced one (c2.jpg,
> recognition stops at certain peak).
> Do you have experience on this?
> Thanks in advance,
> Marcco
Received on Tue Sep 05 2006 - 23:06:00 MEST

This archive was generated by hypermail 2.2.0 : Mon Jan 01 2007 - 16:32:00 MET