[d2n-analysis-talk] LHRS Cross Section Systematic Errors (Update)

[David Flay]

Hi all,

Yesterday (6/14/12) I presented the systematic errors incurred by using a
fit of:

f(x) = (1/x^2)*exp(p0+p1*x)

to model the background signals (positrons or diluted nitrogen) when
subtracting them off from the raw cross section.  I found that the error
is ~1--2% at 4-pass, and 1--8% at 5-pass.  The talk can be found here:

https://hallaweb.jlab.org/wiki/images/3/37/DF_LHRS_6_12_12.pdf

Following up with Brad's comment to see how it adds to the overall
systematic errors, I have computed the in-quadrature sum including these
new errors, and also <excluding> the error estimated for the interpolation
in RADCOR (which was as large as 15% at high Ep) since I now use 44
spectra (in steps of 100 MeV) to unfold our data, which effectively
minimizes the error incurred by the interpolation procedure.

The errors are as follows:

Es = 4730.00 MeV
600	  3.66	2.50	        3.67
800	  3.95	2.50  	3.95
1120	  3.50	2.50  	3.51
1190	  3.40	2.50         3.40
1260	  3.59	2.50  	3.59
1340	  3.59	2.50  	3.59
1420	  3.66	2.50  	3.66
1510	  3.66	2.50  	3.66
1600	  3.72	2.50  	3.79

Es = 5890.00 MeV
600	  3.55	2.12  	3.57
700	  8.69	2.12  	8.70
900	  3.30	2.12  	3.32
1130	  3.68	2.12  	3.68
1200	  3.88	2.12  	3.89
1270	  3.67	2.12  	3.67
1340	  4.25	2.12  	4.25
1420	  4.00	2.12  	4.00
1510	  4.40	2.12  	4.40
1600	  4.00	2.12  	4.00
1700	  6.13	2.12  	6.17

where the first column is Ep (MeV) and the second column is the
'experimental' systematic error (that is, contributions from PID and
acceptance cuts, target density and now the fits);
the third column is that due from radiative corrections (in-quadrature sum
of errors from varying the radiation lengths and the variation of the
F1F209 model).  In contributes a very small amount due to the way it is
added to the experimental error, as shown in my talk at the d2n analysis
workshop.

The errors due to the fit are typically 1% bin-to-bin with the exception
of a few bins, where it escalated to 8% (700 MeV at 5-pass) and a few
other bins where it jumped to 2--5% (which can be seen in my talk above).

Are we satisfied with this estimation, or is another method needed to
asess the fit errors (or do we need a new method entirely to subtract the
background?)

Thanks,

Dave

-------------------------------------------------
David Flay
Physics Department
Temple University
Philadelphia, PA 19122

office: Barton Hall, BA319
phone: (215) 204-1331

e-mail: flay at jlab.org
            flay at temple.edu

website: http://www.jlab.org/~flay
              http://quarks.temple.edu
-------------------------------------------------