[Sane-analysis] d2 using another PDF set

[O. A. Rondon]

I have tried another PDF set, NLO DSSV. There is better agreement
between d2 calculated using g1 and g2 from my empirical A1 fit, and the
values obtained using g1 from the DSSV PDF (with target mass
corrections,) but still d2(3 GeV^2) < 0 with the PDF based g1.

g1p SSFs obtained with the AAC03 and DSSV PDF sets are compared on the
top panels of p. 5 (left at 3 GeV^2, right at 5 GeV^2). The yellow and
purple curves correspond to AAC: in the version that I had been using
(yellow) I may not have included the quark sea properly (purple). In any
case, since the valence contributions are the same in both cases, it
makes little difference for d2, compare the blue points of the middle
plot on p. 5 to those of the top plot on p. 4.

https://userweb.jlab.org/~rondon/analysis/asym/world/d2.pdf

The persistent d2<0 at lower Q^2 using PDF based g1 seems to point to an
issue with all PDF sets, which probably are not reliable below Q^2 ~ 4
GeV^2, for x^2 > ~ 0.2 (x ~ 0.45). Since the twist-3 contribution is
more important at 3 GeV^2, using fits to A1 and A2 data does seem to be
a better way of getting d2. In any case, we should be wary of what data
sets are included in the PDF analysis, because they could be biasing the
  distributions at low Q^2 and high x.

Hopefully, between the UVA-JLab analysis and the Temple analysis we'll
have consistent data that can be reliably combined with SLAC's true A1
and A2 for improved fits.

Merry Christmas, Happy Holidays, Happy New Year,

Oscar