H

1
Update on Signal Significance in H?gg
2
Motivation and Outline

• Motivation
• Study of significance from fitting can see effect
  from
• signal shape, effect due to floating higgs mass.
• Combined significance summed in quadrature is not
  necessarily accurate.
• Work done since last meeting
• update pseudo data according to samples for csc
  note.
• new MC_at_NLO and EWaajj included.
• Study inclusive, Hjets.
• Add H0jet for combined fitting (H0,1,2jets).
• Study the higgs mass floating effect and fitting
  stability.

3
Significance calculation with likelihood fit

• Compute LR distributions from Toy Monte Carlo
• Randomly generate a pseudo-experimental outcome.
• Perform fits (RooFit) with SB hypothesis and B
  (Background-only) hypothesis
• compute ln(LR)ln(LSB/LB)
• Also called ?NLL
• Repeat many times to get a probability
  distribution
• Compute two probability distributions
• for outcomes with signal (SB toy MC) and
• For outcomes without signal (BG-only toy MC)
• Compute CLB from plots ?NLL(BG-only toy MC) and
  ?NLL(SB toy MC) and convert it into
  significance.
• Samples
• 10 fb-1 inclusive (Ns438,Nbkg76021).
  MH120GeV.
• 10 fb-1 H ?gg1jet (Ns65,Nbkg5633). MH120GeV.
• 10 fb-1 H?gg2jets(Ns14,Nbkg148). MH120GeV
• Mgg window from 110 to 150GeV.

4
Fit procedure

• Fit formulae ME(MH)NSS(MH)NBB(MH)
• S(MH) is a Gaussian (mass allowed to float).
  B(MH) is linear (for H1jet and H2jet case,
  exponential for H0jet).
• Nsgt0.
• For Bkg hypothesis fit fix Ns0.
• For combined case H0jet ,H1jet and H2jets are
  considered as three categories, fit
  simultaneously with same signal mass.

5
Notation for the fit
Two types of fits BG hypothesis fit, SB
hypothesis fit, two types of toyMC BG-only
toyMC, SBG toyMC. 4 possibilities are following

BfitB BG hypothesis fit BG-only toy
MC BfitSB BG hypothesis fit SBG toy
MC SBfitB SB hypothesis fit BG-only
toy MC SBfitSB SB hypothesis fit SBG
toy MC At the left side of fit is the type of
fit. At the right side of fit is toyMC
6
One example from inclusive analysis
MH120GeV 10fb-1

MC SBG toyMC (left column), BG-only toyMC (
right column). Fit BG-only hypothesis(at the
top) SBG hypothesis (at the bottom)
7
Fitting result and calculation of significance
(inclusive)
MH120GeV 10fb-1
1. Compute CLb
inclusive
from Bkgs ?NLL distribution

2. Significance can be got from Erfc(S/sqrt(2))
CLb .
S 3.3 for inclusive case Event counting 3.8
?NLL0
?NLL0 is defined as median of the SB distribution
8
Floating higgs mass effect
inclusive 10 fb-1
MH allowed to float
MH fixed at 120GeV
?NLL0
?NLL0

• Compare the left plot with the right one
• the floating higgs mass doesnt affect a lot
  the fittings on SB data.(black).
• But it truncates the tail of the fittings on
  B data around a factor of 10.
• The significance MH allowed to float (3.3),
  MH fixed(4.0) (statistics is limited) .
• Event counting (3.8).

9
Fitting result and significance for H1jet
MH120GeV 10fb-1
MH allowed to float
MH fixed at 120GeV
Significance 1.9 (MH floated), 2.4 (MH fixed)
10
Combination of H0,1,2jets
MH allowed to float
S 4.0 (statistics is limited). Lack of CPU
time to run fixed Higgs mass case.
11
Stabilities of the fits (MH is allowed to float)
Percent variation of significance due to the
following variation H0jet Bkg shape fixed
10 Ns varies 10 12/-12 Nbkg
varies 20 -9/13 Nbkg varies 40
-18/27 H1jet Bkg shape fixed 12 Ns
varies 10 10/-10 Nbkg varies 20
-8/15 Nbkg varies 40 -14/32 H2jets Bk
g shape fixed 11 Ns varies 10
10/-10 Nbkg varies 20 -8/13 Nbkg
varies 40 -15/26

• Fixing Bkg shape can cause roughly 10 variation
  to significance
• Although the significances are different from
  s/sqrt(b),
• the variations due to Ns, Nbkg changes are
  consistent with it.

12
Conclusion and Plan

• Calculations of signal significance with
  likelihood fits for inclusive, Hjets and
  combination of them have been done.
• Floating higgs mass effect and stabilities of
  the fits have been studied.
• the former has significant effect on the fitting
  result.
• Systematics of the fitting needs to be studied.
• Plan try more sensitive ways to evaluate signal
  significance.

13
Backup slides
14
Procedure for Combined analysis