Hi,
I am not sure but I find the conclusion in Doug's "handout" make very good
sense. It is essentially the same as the "more natural approach" in
Michael's summary last week if I understood both right. I am not sure
where the confusion still is, but maybe the following line of unrigorous
thought is helpful.
For both GeN and T20, we meausre one thing only. We don't really care
about the raw asymmetry or if the polarization are the same. So divide the
raw asymmetry by the polarization and pretend both are from experiments
with 100% polarized target. Then the combined error should scale as
1/sqrt(N1+N2) and there is a very straight forward way to achieve this
behavior from 1/sqrt(N1) and 1/sqrt(N2). :)
So long as one don't go into the shady area of combining systematics, I am
not sure why so much discussion on this. In regarding to GeN and T20 cos
we do have observables independent of the run conditions. I can imagine
other channels might be a bit complicated cos the final product one pulls
out depends on the spin angle setting etc so it is probably illegal to
just take the average.
Michael pointed out the "usual approach" of combining the counts is
incorrect. In fact I am wondering when and why that became the "usual
approach" :)
I am not sure I understand the relation between the "running six
experiments" arguement and the issue of combining
different-polarization-data. So long as I can see, we have only TWO major
sets of data, one last July-Oct, one this year, polarizations of these two
sets are different. If hPz/Pzz are different within the group of the SIX
experiments, we might as well go to Idaho and make a living growing
potatoes.
My 2 cents. Now I understand why Doug's write up is so long. :)
Chi
On Sun, 22 May 2005, Douglas Hasell wrote:
> Hi,
>
> Apologies to anyone who actually read my note on combining
> asymmetries.
>
> Vitaly has pointed out that we do not make simultaneous
> measurements of the asymmetry (except when we do left-right
> asymmetries). Rather we do separate measurements for beam and target
> spin combinations and normalise to charge. This means the statistics
> are Poisson and not correlated
>
> I will rework the note but I don't think things will factor out
> as nicely now which means combining measurements will be tedious.
>
>
> Cheers,
>
> Douglas
>
> 26-415
> M.I.T. Tel:
> +1 (617) 258-7199
> 77 Massachusetts Avenue Fax: +1
> (617) 258-5440
> Cambridge, MA 02139, USA E-mail:
> hasell@mit.edu
>
>
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:32 EST