[Halld-tagger] [EXTERNAL] Fwd: Accidental subtraction

[Richard Jones]

Hello Peter,

Thanks for asking, I noticed this as well, but I thought it was only being
used for PWA where the primary focus is on the angular distributions. I
agree that it is a concern for differential cross sections. You are right
that this is not properly taking into account the accidentals that are
present. Essentially it amounts to a hybrid between a fully tagged and a
fully untagged experiment. Here are those two extremes:

   1. A(untagged) -- the photon energy is inferred from the reconstructed
   final state, and used to compute all of the beam properties associated with
   the event: the flux, the polarization, etc.
   2. B(fully tagged) -- the photon energy is inferred from the
   unambiguously associated hit in the tagger, which is used as input to the
   kinematic fit and to lookup beam properties for the event

At the rates of GlueX phase 2, we do not have the luxury of option 2 on an
event-by-event basis, but we can achieve it by accidentals subtraction.
Short of full accidentals subtraction there are several short-cuts you can
use. All of these have uncontrolled systematics.

   1. best chi square - put them all in a ring and take the last man
   standing as the winner with weight 1
   2. weighted average - count them all above some chi-square acceptance
   cut and weight each event by 1/n where n is the number of surviving tags

Both of these methods reduce to tagging strategy A(tagged) at low rate,
while they reduce to strategy B(untagged) at high rate. At GlueX Phase II
intensities we are some intermediate hybrid of the two with these
shortcuts, certainly not approximating B(fully tagged).

To see what these short-cuts entail, consider the high-rate limit in the
tagger. At high rate, the extracted cross section goes to infinity for a
realistic tagger and an ideal GlueX detector. In reality, the asymptote
would be something greater than one, channel and final state dependent, and
probably run period dependent as well. The reason for this is that the
tagger detection efficiency per beam photon goes down at high rate, while
the accidentals continue to grow and generate a valid result for any
reconstructed final state, tagged or untagged. So the flux that you need to
put into the denominator under the yield for extracting a cross section
will be different depending on the final state. Using the same flux
regardless of final state could be a leading cause for why we are seeing
different cross sections for the charged and neutral decays of eta.

Beyond that, the shape of the flux spectrum (and the polarization spectrum
for polar observables) is different from the shape of the energy dependence
of the reconstructed yield. Take for example the energy-dependent cross
section around the coherent edge. The tagged flux has a sharp edge, whereas
the reconstructed yield washes out the edge with a resolution that depends
on everything in sight: the kinematic fit cut, conditions in the detector,
the particular final state, etc. One way to reduce our dependence on the
different beam photon energy resolutions in the flux and yields is to
average over a wide range in beam energy. As long as we are not interested
in the s-dependence of the cross section, this might be justified and would
reduce the systematics from these short-cut approaches.

In PWA, i understand that this avoids the pain of negative weights and so
improves the statistical error from the fits (or at least it gives that
feeling). In fact, it introduces a set of new systematic errors of its own
that will probably drive us back to the more rigorous approach before we
are done. For the moment I am not speaking up about this because we just
need to get our first results out. But eventually this needs to be given a
critical review. I hope to be part of that at some level, as soon as my
work on photon beam systematics reaches a level where it can be used for
publications.

-Richard Jones

On Sun, Feb 19, 2023 at 11:13 AM Peter Hurck <Peter.Hurck at glasgow.ac.uk>
wrote:

> *Message sent from a system outside of UConn.*
>
> Hi Richard,
>
> During the collaboration meeting a few people presented analyses which
> used a chi^2 ranking with a +-2ns cut around the RF peak instead of tagger
> accidental subtraction.
>
> My initial thought was that this is wrong and not recommended. Did the
> guidance by the beam line group change regarding this issue? Given that the
> current a2 cross-section analysis is using this method and there is a big
> push to publish it asap I am concerned that this might not be resolved
> properly and might set a bad precedent going forward.
>
> Given that you are the expert on this topic, what are your thoughts on
> this issue? Is that a legitimate way to perform analyses?
>
> Cheers,
> Peter
>
>
> ----------------------------------------------------
> Dr Peter Hurck (né Pauli)
>
> My new email address is
> Peter.Hurck at glasgow.ac.uk
>
> Research Associate
> Nuclear and Hadron Physics Research
> School of Physics and Astronomy
> University of Glasgow
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-tagger/attachments/20230220/f23fa628/attachment.html>