1 / 9

Oversmearing Tunfold Errors

Oversmearing Tunfold Errors. Michael Gardner Dilepton – Z Group May 26 th , 2014. Reminder. Comments on MC vs. Data: Z-> ee in PbPb plot. ~350 Zs in ee in PbPb Data – given these stats, and the fact that yield is based on counts (not fits), want to justify not doing more calibration.

jeb
Download Presentation

Oversmearing Tunfold Errors

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. OversmearingTunfold Errors Michael Gardner Dilepton – Z Group May 26th, 2014

  2. Reminder • Comments on MC vs. Data: Z->ee in PbPb plot. • ~350 Zs in ee in PbPb Data – given these stats, and the fact that yield is based on counts (not fits), want to justify not doing more calibration

  3. Kolmogorov-Smirnov Test • Results of K-S test listed here are uniform, from 0 to 1. • For Zee in PbPb: • 30 bins, K-S result: 0.239 • 60 bins, K-S result: 0.165 • UnbinnedK-S result: 0.061 • 30 bins (60-120), Χ2 p-value: 4.0348e-007 (tails). • 30 bins (80-100), Χ2 p-value: 0.121982. • Not Terrible: we try oversmearing MC, to see what the best fit is, and see effect on acc, eff, etc.

  4. Oversmearing • Idea: • take each electron (or muon), and smear it’s pT (for each lepton, I did this 20 times): • 1. adding a random number from a gaussian distribution of mean 0, with sigma = M GeV (modeling error in background subtraction). • 2. multiplying by a random number from a gaussian of mean 1, with sigma = N% (modeling error in reconstruction). • recombine electrons to form new Z. • compare new distributions, find smearing that gives smallest Χ2 (looking 80 < Mass < 100).

  5. Results • Z  ee in PbPb: • Χ2 p-value Plot: • Value lower than shownbefore (6%), since MCdone20x. Going forshapenot exact value. • Adding: 1.8 (1.6) GeV shift (Χ2/ndf = 0.84);  should be looking at 10 bins? 1.5 GeV • Multiplying: 4.5% shift (Χ2/ndf = 0.90); • No Smearing: Best Smearing:

  6. Results for All • Z ee in PbPb: • Adding: 1.5 GeV shift (Χ2/ndf = 0.84); • Multiplying: 4.0% shift (Χ2/ndf = 0.90); • Z ee in pp: (MC is slightly wider) • Adding: 0.0 GeV shift (Χ2/ndf = 1.16); • Multiplying: 0.0% shift (Χ2/ndf = 1.16); • Z μμ in PbPb: • Adding: 0.7 GeV shift (Χ2/ndf = 0.94); • Multiplying: 1.5% shift (Χ2/ndf = 0.92); • Z μμ in pp: • Adding: 0.6 GeV shift (Χ2/ndf = 0.67); • Multiplying: 1.5% shift (Χ2/ndf = 0.65);

  7. Effect of Smearing on Acc * Eff • Most heavily seen vs. pT. In centrality there will be a small decrease in efficiency, as those Zs close to the 60 and 120 GeV boundaries may be smeared out of the range on not be counted. • For Z  ee in PbPb, with the 1.8 GeV shift: • Overall Effect: Acc x Eff drop of 0.02%. • Big change vs. Centrality: 0.1% -. • vs. y: 0.8% -, 0.7% +. • vs. pT (first 2 bins 0-5, 5-10): 13% -, 17% +.

  8. TUnfold Errors • Problem #1: Stat. Errors, not propogated through Tunfold. • Z  ee in PbPb: • For TUnfold, create a Matrix of Gen vs. RecopT • pT bins (for Tunfoldnum_rec > num_gen): • RecopT Bins = 0;5;10;15;20;25;30;40;50;75;100;500. • Final pT Bins = 0;5;10;20;30;40;50;100. • Running 100k Toy MC, the Stat. Uncertainties: • Before TUnfold: • Num_pT[nBins] = {94,91,52,25,21,11,16,12,5,2,0} • StatUnc % ~ {10,10,14,20,22,30,25,29,45,70,…} • After: • Final_num[nBins] = {100.4,90.3,64,35,14,18,8} • Fin.StatUnc % = {14,18,17,23,44,34,40} • What was Used = {10,10,11,18,25,29,38}

  9. Summary of Z values • https://twiki.cern.ch/twiki/pub/CMS/DileptonEWK/Summary_Values.txt

More Related