[Clascomment] OPT-IN:Target and Double Spin Asymmetries of Deeply Virtual Ï0 Production with a Longitudinally Polarized Proton Target and CLAS

[Andrey]

Line 172: The constant dilution is in fact the parametrizetion. The low 
statistics doesn't allow to use ratio for each bin individually. So we 
fit it with constant function (the error bars are too large for any 
other complicated model to be used for fitting). I didn't list the 
factors because we have 3 of them for each ECEC, ICIC and ECIC 
topologies. I thought if I include 3 numbers I should better explain why 
we have different dilutions for different topologies and decided to 
leave it for a longer paper. I can include them if you think it is 
absolutely necessary.

Line 220-223: 3% is actually for PbPt value (for DSA) taken from elastic 
measurements, we use statistical error from the elastic analysis to 
evaluate the systematic effect of this PbPt value on the asymmetry.  In 
case of target spin asymmetry the Pb error comes into play, you are 
right. Still, we have error in numerator 3% and in denumenator - 2%. The 
resulting error is slightly above 3%, if I am not mistaken. However, the 
error on Pb is quite conservative. And this particular source of 
systematic uncertainty is relative, not absolute. So its contribution is 
really negligible comparing to other sources.

Fig. 3 - The bumps are mostly for the bins with lowest statistics. I 
used the method (described in Analysis note and discussed during eg1dvcs 
meetings) to negate the statistical effect on systematic uncertainty as 
much as possible. But with event selection procedure (which usually 
contributes the most to the systematics) it is very hard to remove it 
completely.

Lines 244-256: Yes, I tested the fits with AUUcos1 and AUUcos2 fixed to 
zero and also with free AUUcos1 and AUUcos2. The numerator terms are 
fairly insensitive to these changes (often observed in previous 
asymmetry measurements as well). I compared the AUU2 and ALU with 
previous experimental measurements and they agree reasonably well (in 
the analysis note). We used it these comparison to cross-check our analysis.

General Remarks: We can explore different binning in the long paper.

Fig. 4: Yes, this systematic band is for fitted parameters, so there is 
all the uncertainties that come from the fit. 4.5% average listed in 
paper is actually for the systematics measured per {Q2,xb,-t,phi} bins 
(before fits). I assume these results are the most fundamental ones, 
since with their knowledge the fitting can be redone and/or improved in 
the future with theoretical model developments.

I'll fix typos, thank you.
Andrey.


On 09/27/2015 04:02 PM, Sebastian Kuhn wrote:
> Line 172: You say "a constant dilution factor was applied". Why not use the measured ratio between the total number of events in the peak and the extrapolated 12C background, separately for each bin? Or at least a parametrization thereof? At least, you should quote the value of the "constant" dilution factor used and give some estimate how much it could vary from bin to bin.
> Line 215: "from 2.5sigma TO 3.5 sigma".
> Lines 220-223: 3% seems quite small, given that Pb alone already has an uncertainty of 2%/84% = 2.4%. What is the uncertainty in PbPt? (Not only the count rate uncertainty from elastic scattering, but also systematic uncertainty from kinematics, form factor models, dilution in the elastic channels etc.)?
> Fig. 3 caption: rows and columns are switched. Also, the systematic uncertainty bands show big bumps that must be larger than 4.5% -> explanation?
> Lines 244-256: I understand the AUUcosphi is very hard to constrain, but did you vary it within expected range and check the corresponding systematic uncertainties for the extracted AUL and ALL moments? Also, did you use any constraints from the known ALU and AUUcos2phi? Do you get similar results?
> GENERAL REMARK: Only 2 bins in x and Q2 may be a bit coarse. Most of the results seem to have small enough error bars to allow up to 4 bins (smaller bins in x). Just a suggestion - you be the judge.
> Fig. 4: Explain more carefully the systematic uncertainty bands. They look bigger than the quoted 4.5%, so I assume they do contain some uncertainty from the fit? (i.e., cross talk between numerator and denominator)
> Line 278: "GL" -> "GGL"