Parametrisation dependence of the fits to new data

1
Parametrisation dependence of the fits to new
data April 3rd 2009 A M Cooper-Sarkar

• Treating errors as 110 quadratic and 3 OFFSET
  using correlations gives similar results.
• I have been investigating parametrisation
  uncertainty in the RTVFN fit using 1702 QCDNUM
• There is also progress in agreement with Voica
  with 1702 and ZMVFN
• And STOP PRESS at 0008 today we have agreement
  on RTVFN

2

• Variations of the parametrisation

In pink the central fit and 3 variants which
matter
The central parametrisation has the basic 9
parameters of the in-between PLUS Euv The ?2 is
576 for 592 data points. This is not the very
lowest obtainable but was chosen because of the
staibility of the fit wrt error treatment
(correlations) AND because the high-x valence/sea
shapes are reasonable I have tried 4 other 10
parameter fits- none of them competes in ?2
However if we move to 11 parameter fits then
some do
For the last 4 the extra parameters are 0, but
for the first 3 there are some significant
differences. I have used these 7 fits to estimate
parametrisation uncertainty
3
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits Valence is affected for all-x Sea
and gluon only at higher-x
Just to remind you- this is what the model
uncertainties look like on uv,dv,Sea, glue
4
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits
And this is what the parametrisation
uncertainties look like using only the first 3 of
the 11-param fits- clearly we need only consider
these 3
5
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits Sea is affected at higher-x
Just to remind you- this is what the model
uncertainties look like on ubar,dbar,cbar,sbar
6
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits Valence is affected for all-x Sea
and gluon only at higher-x
Just to remind you- this is what the model
uncertainties look like on uv,dv,Sea, glue
summary plot
7
parametrisation
models
Alternative ways to present the information at
high-x
parametrisation
8
Agreement with Voica on QCDNUM17 Quadratic
errors, old chisq, ZMVFN
Voica
Mandy
FCN 560.6832
560.651697 1 Bg
0.30660 0.32400E-01
0.305753296 0.0344224902 2 Cg
9.1161 0.66451
9.08992953 0.727072239 3 Buv
0.70890 0.20548E-01 0.707941564
0.0214990681 4 Cuv 5.0389
0.22601 5.04514589
0.250542822 5 Duv 0.0000
fixed 6 Cdv 4.3939 0.32248
4.38026475 0.374606914 7
Adbar 0.11674 0.42248E-02
0.1169537 0.00468142 8 Bdbar
-0.20930 0.48790E-02 -0.209200986
0.0053179119 9 CDbar 3.0429
0.74298 3.07734298 0.923174655
10 CUbar 1.6954 0.26912
1.69344075 0.29576455 11 Euv
10.162 2.1751
10.2500163 2.37210416 12 Dg
0.0000 fixed 13 Ddv 0.0000
fixed 14 alphas 0.11760
fixed
9
Agreement with Voica on QCDNUM17 Quadratic
errors, old chisq, RTVFN at 0008 am
Voica
Mandy

• FCN 546.036
  546.055
• 1 Bg 0.19247 0.30484E-01
  0.192822802 0.0334332937
• 2 Cg 8.2850 0.66604
  8.29443626 0.772864482
• 3 Buv 0.69687 0.27545E-01
  0.696943746 0.0291369056
• 4 Cuv 4.7163 0.13958
  4.7159869 0.144654663
• 5 Duv 0.0000 fixed
• 6 Cdv 3.8240 0.30430
  3.8233966 0.380288017
• 7 Adbar 0.17497 0.72288E-02
  0.175143 0.0084316568
• 8 Bdbar -0.15575 0.55338E-02
  -0.155718978 0.00634262042
• 9 CDbar 3.3633 0.75800
  3.36345492 1.0071895
• 10 CUbar 3.6443 0.51168
  3.6483837 0.561816451
• 11 Euv 9.9063 1.7546
  9.89503835 1.81193827

10
Summary

• Parametrisation uncertainties investigated
• So far 4 out of 4 other 10 parameter fits do not
  compete
• 3 out of 7 possible 11 parameter fits make some
  difference
• Parameter space not yet exhausted but gives a
  good idea- high-x is affected AND valence quarks
  BUT the beauty of low-x gluon and sea is not
  affceted
• Could investigate more variants x3 and vx terms
• BUT perhaps all possible variants need not be
  ready for DIS?
• Agreement with Voica HOORAY
• What do we do for DIS?

11
extras
12
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits
And this is what the parametrisation
uncertainties look like using only the first 3 of
the 11-param fits- clearly we need only consider
these 3
13
And this is what the parametrisation
uncertainties look like using 7 different
11-param fits Valence is affected for all-x Sea
and gluon only at higher-x
And this is what the parametrisation
uncertainties look like using only the first 3 of
the 11-param fits- clearly we need only consider
these 3