Download presentation
Presentation is loading. Please wait.
1
Photon energy smearing from events
Glen Cowan, Sudan Paramesvaran, David Hopkins, Henning Flaecher Royal Holloway, University of London EMC Calibration Meeting 6 Sept 2006
2
Context Follows on from resolution studies by David, Henning, Sudan.
Same mmg n-tuples and selection as before, see e.g. David Hopkins talk 5 April 06 at EMC software meeting: 2 ‘good’ measured tracks 1 ‘good’ measured photon 1 identified m (loose) Kinematic fit c2 prob > 0.05 184 fb-1 → 1.2 million events
3
Goal of study For the photon in the mmg events we have
Emeas measured by the calorimeter Efit from the kinematic fit Histograms of x = Emeas/Efit found to have high-x tail, especially for higher Eg Goal: (Roodman, Kocian, ...) try smearing MC photon energies to give better data/MC agreement.
4
Method (1) Scale to same area as data histogram
Try to smear MC so that it looks like data, i.e., find a pdf s(z) such that y + z ~ f(x) Find For x = y + z
5
Method (2) For x = y + z we have In terms of the histograms this is
where Find For a parameterized pdf s(z;q ) we therefore have original MC smeared MC smearing matrix
6
Method (3) Try for smearing pdf: Gaussian Student’s t
For Student’s t, n controls extent of tails n = ∞ is Gaussian, n = 1 is Cauchy Use binned ML but for now ignore MC statistical errors, equivalent to minimizing
7
Gaussian fit Student’s t fit
8
Gaussian fit Student’s t fit
9
Gaussian fit Student’s t fit
10
Gaussian fit Student’s t fit
11
Gaussian fit Student’s t fit
12
Gaussian fit Student’s t fit
13
Gaussian fit Student’s t fit
14
Gaussian fit Student’s t fit
15
Gaussian fit Student’s t fit
16
Gaussian fit Student’s t fit
17
Gaussian fit Student’s t fit
18
Gaussian fit Student’s t fit
19
Gaussian fit Student’s t fit
20
Gaussian fit Student’s t fit
21
Gaussian fit parameters
22
Student’s t fit parameters
For 1.0 < E < 1.5 GeV, →∞ (consistent with Gaussian).
23
Andrew Wagner, 29 Aug 06 Neutrals meeting, showed
similar study with L/R asymmetric Gaussian smearing:
24
Andrew Wagner, 29 Aug 06 Neutrals meeting:
25
From here, Parameterize the fitted parameters vs energy, (e.g. polynomial), then for each photon generate z ~ s(z;q(E)) and replace E → E (1 + z) Resulting distribution will not be exactly equal to smeared MC from fit due to several approximations made, but should be close. To what extent are the tails due solely to EMC response? Are they in part caused by, e.g., modeling of tracking, backgrounds,...? Some further steps: investigate angular dependence, refine E binning, need more MC data (also use more real data?), check effect on other quantities, e.g., p0 peak
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.