[Clascomment] comments on The Beam-Target E asymmetry for vec gamma vec n -> pi- p in the Nâ resonance region

[Dr. A.M. Sandorfi]

Dear Dave,

Thanks for your two comments.

(1) Regarding the PWA "bands" in figure 4, there are a variety of ways of
showing that there is some spread inherent to PWA predictions that changes
with kinematics. Eugene had a similar question, and I'll repeat part of my
response: The curves are from Energy-Dependent (ED) PWA. In the case of
SAID, these are smooth parameterizations of a set of Energy-Independent (EI)
solutions. The latter are made by binning the world's data in W and
assigning all measurements within a particular bin to the centroid energy of
that bin, even though all those data were not taken at exactly the same
energy. So there is an inherent fuzziness in this PWA process. (BnGa does
the reverse: they fit parameters of an ED PWA, and then check consistency
with an EI solution, constructed in the same way as SAID.)

So far as I am concerned, both the new SAID and the new BnGa PWA provide
very good representations of the asymmetry data. There is no reason (or
attempt in the paper) to favor one over the other. I am very reluctant to
add another band to figure 4, for fear of obscuring the data itself among a
maze of curves and bands. (This is a different situation from your KL paper.
Apart from having more space and larger figures, you had only one solution
that was actually refitted to your KL data. As a result the different PWA
bands in your figures were naturally separated. In our case, both SAID and
BnGa were refitted to the E data and so lie on top of one another.)

To help alleviate concerns over any perceived bias between PWA solutions, I
have changed the text to indicate that the SAID band is used just for
illustration of the kinematic dependence. That paragraph now reads:

"   New Partial Wave Analyses (PWA) of $\pi$ photoproduction have been
carried out, augmenting the neutron data base with these new $E$
asymmetries. New PWA from the George Washington University Data-Analysis
group (SAID) \cite{SAID}, and new PWA from the Bonn-Gatchina (BnGa) group
\cite{BnGa}, are shown as solid red and solid black curves in
Fig.~\ref{E_g14wPWA}, respectively. Both provide very good representations
of the new $E$ data. PWA combine results from many experiments at different
energies, and this results in varying degrees of sensitivity to energy and
angle. This is illustrated by the red bands whose width indicates the SAID
variation across the energy bin.   "

(2) As to combining the three different analyses, while we use the weighted
mean as the best estimate of $E$ at each kinematic setting, the associated
error is certainly NOT the standard error on a weighted mean. This situation
is common in High Energy where multiple analyses are carried out on
basically the same data set, and we have used their methods to estimate the
correlations and calculate the associated uncertainties - see ref [22] by
Schmelling. This is discussed in some detail in chapter 5 of our Analysis
Note. Numerically, the net uncertainties always come out to be bounded by
the largest and the smallest errors of the three analyses. (So no, this is
not just using the same data three times over.)
   In CLAS physics it's rare to have more than one independent analysis, but
when there is, combining them has two advantages:
(a) Every analysis method has some systematic bias that, since it's source
cannot be identified, remains uncorrected and lumped into the quoted
experimental systematic uncertainty. (eg. Although the three sets of
analyses are statistically consistent, nonetheless the red points in figure
3-left panel (W=1580) tend to be the lowest (more negative) of the three
analyses, at least at that energy.) Averaging over three completely
different methods gives a better "best estimate" for the central value.
(b) The acceptance at very forward and backward angles is different for the
three methods and combining all three gives a more reliable value at angles
where interference effects between different multipoles are large.

Hope this clarifies your points.
All the best,...Andy





On 3/16/17, 8:34 AM, "David Ireland" <David.Ireland at glasgow.ac.uk> wrote:

> Dear Andy, et al.
> 
> Congratulations on getting this analysis to the submission stage. Looks like a
> nice set of results, which have the power to improve the theoretical models.
> 
> I have a couple of comments:
> 1. In our paper last year on KY observables, we provided Bonn-Gatchina bands
> for the calculations similar to what you have done with SAID, where in our
> case the bands indicated the spread in the event-weighted observable
> calculation. However, for the comparison between SAID and BnGa calculations to
> be meaningful they should either both be single lines or both bands.
> 
> 2. In para 1 of the RHS of page 3 you say "A weighted average of the results
> from the three analyses has been used as the best estimate of the $\pi^- p$ E
> asymmetries." I may have interpreted this incorrectly, but with three
> different analyses you can only compare them with each other to establish
> consistency. You must then pick one method for final results; using an average
> amounts to using the data three times over, which is not usually allowed in
> statistical analyses.
> 
> Best wishes,
> 
> Dave
> 
>