Bugs found

Two bugs have been discovered in the last few days, both of which were introduced in early Dec.  At that time, the code was being updated to cover the cases needed for an eta-dependent cross section and for the A_N measurement.  I also altered my file naming convention to keep track of more information than it had previously.

Bug 1: Dec 4th

The (SMD u+v cluster energy)/(photon energy) cut was erroneously set to 0.012 instead of 0.006 for the data. The cut was still left at 0.006 for the Monte Carlo efficiency calculation.  Jim commented that things looked different, the mistaken value was found, and new files were generated on Dec 6th.  Unfortunatly, the new files were not copied to correct directory, and all of my cross section and A_LL results used the data files with the wrong cut, starting Dec 4th.  Since the reconstruction efficiency didn't have this cut included, the result was a decrease in the cross section.


Bug 2: Dec 7th

The code for computing the BFC and Pythia filter efficiencies and the code for computing the trigger efficiency (within the filters) was updated to allow for either 2D (eta, p_T),or 1D eta or 1D p_T.  The fundamentail histograms are made in 2D, and then projections are made before taking the ratio.  A feature of ROOT, that I was not aware of at the time, is that if one calls TH2::ProjectionX("newName") with no specification of the bin range to use, then the overflow bins are included.  Thus, when I made the p_T projection, ROOT effectively removed my cut on physical eta between 0.8 and 2.0.  This had the effect of decreasing the efficiency (especially for the Pythia filter), and thus increasing the cross section.

The increase due to this bug almost canceled the decrease due to the first bug.  The two bugs thus masked each other, and were not found until January, when it was adjusting cuts and found that I couldn't reproduce the results from December.

Results Before the Bug:

These are links to my blog from Nov 7th, the first time I showed the cross section with the inclusion of the BFC filter efficiency and the last posted cross section before the bugs were introduced.


Reproducing the old results:

There have been minor updates to the background subraction method (such as automatically rebinning if there are small statistics) and the plotting code (such as the method to interpolate the value of the theory curve at a given pT) which cause almost negligable effects, except at the highest pT where the rebinning can account for the change.  This is with the MC L2 thresholds set to the identical value as the data--label 'h' according to this scheme.



Here's the same plot with one more bin on the left side:


Updated trigger threshold

It was noted in this blog that the MC threshold for the L0 needs raised 3% above that for data.  The plots on that blog are biased due to the above bug, so here is the updated version of the cross section with the 3% higher L2 threshold--level 'i' according to this scheme.  Note, the data is still at level 'h'.


The cross section has increased for all bins, though only in the 5-6 GeV and 12-16 GeV is the effect noticiable.

Conclusions

Bugs found and fixed, and these are now the most up-to-date plots for the cross section.