Study switch from ElectronVeto to pixelseed

[jdgrassi]

pixelseed is pico branch currently, can be studied in draw_pico. Once study is finished, either update nano2pico to use pixelseed, or justify choice of electron_veto in future presentations.

[andersbarzdukas]

NOTE: everything below has eVeto applied already.

Look at photon_pixelseed variable

Looks like this change may be useful. attached are plots htozgamma_kingscanyon_v1 picos with the overlap between photons and electrons removed (via photon_elidx branch).

Skeletonll_TIGHT_BASELINE_pixel_seed_so__photon_pixelseed0wgt_1invfb__lumi_nonorm_lin.pdf Skeletonll_TIGHT_BASELINE_pixel_seed_ls__photon_pixelseed0wgt_1invfb__lumi_shapes_lin.pdf Skeletonll_TIGHT_BASELINE_pixel_seedphoton_pixelseed0__wgt_1invfb__shapes_lin.pdf

Impact on mllgamma

The plots are of the photon_pixelseed variable which is 0 if the photon has no pixelseed (good photon) and 1 if the photon has a pixel seed (bad photon). The first three plots show have the y-axis set to % of events and the rest have the y-axis set to event yields. Note: the plots have signal multiplied by 10, but this is not correctly labeled on the plots.

Events with photon_pixelseed==0 m_eegamma with photon_pixelseed==0 m_mumugamma with photon_pixelseed==0 m_llgamma with photon_pixelseed==0

Events with photon_pixelseed==1 m_eegamma with photon_pixelseed==1 m_mumugamma with photon_pixelseed==1 m_llgamma with photon_pixelseed==1

It does seem that events with photon_pixelseed[0]==1 have the bkg plateau near the peak which may make the overall shape easier to fit?

Event yields:

Requiring pixel seed ==0 removes ~5% of signal MC events and ~20% of fake photon events. The table below contains the yields where pixelseed ==0 and ==1.

table_pixelseedcheck.pdf

The S/root(B) and separation of just the higgs window were added to the table. The S/root(B) in the higgs window is significantly lower when there is a pixel seed, so it seems like any losses would be negligible. The signal efficiency hit would be 5% and the background efficiency is 91.3% within the higgs window. The epsilon_S/root(epsilon_B) is 0.994, so based off of this metric it seems like the improvement may be minimal.

This will be part of a future presentation, and I can show this in a Thursday meeting.

[mcoshiro]

Also worth noting: if this switch is made, the SFs for the electron veto would need to be switched to the pixel seed veto SFs, which are also centrally provided