Update on Hist vs Func Templates
In my talk for today's (Feb 7, 2013) Spin PWG phone meeting, it was noted and discussed that the uncertainty for the longitudinal spin asymmetries is much higher using histogram rather than functional templates. One additional point has been noted: alpha (the data/MC energy scale factor) was not fixed for the spin fits as it should. Here's updated results (plus a whole lot more plots).
Usual Freeing/Fixing of Alpha
Typically, the fits progress as follows:- Fit the data set used for the cross section, initializing the weights to the value in the MC and setting alpha to 0.97
- Fit the longitudinal data set, summed over spin states, with weights and alpha initialized to the value from (1), with alpha still free
- Fit the longitudinal data set, seperately for each spin states, with weights and alpha initialized to the value from (2), with alpha fixed
Results
Results are attached at the bottom of this blog, organized in the same fashion as the last several posts: first are the mass and fits per pT bin, starting with 5 < pT < 6. The peak position is also plotted, and for the "tot" file, the longitudinal spin asymmetries are also given. Note: only the results for the histogram fit are attached, as the comparable functional fits are given in the Feb 7, "Repeat of Cut Review" blog, with file names including "LONG" and "oldCuts2-oldCuts2".Conclusions
We see that the "blow up" of uncertainties is limited to the "background subtraction before asymmetry" estimate of the background. For clarity on "before" vs. "after", see slide 4 and 5 of my Jan 10th Spin PWG phone conference talk. In these fits, alpha is fixed (though still having a fixed uncertainty), so it is expected that the increased uncertainty is not due to the uncertainty on alpha. It does appear that the uncertainty on the sum of the histograms (given in green) is quite high, which would thus drive up the uncertainty on the weights being determined in the fits.Groups:
- sgliske's blog
- Login or register to post comments