csc tth(hbb) further analysis

17
1 CSC ttH(Hbb) further analysis Chris Collins-Tooth 12/11/2008

Upload: raina

Post on 18-Mar-2016

46 views

Category:

Documents


2 download

DESCRIPTION

CSC ttH(Hbb) further analysis. Chris Collins-Tooth 12/11/2008. Motivation. Some late changes were made to the CSC note “cut-based” analysis. I felt that I didn’t have time to look at the effects on background shape (which were especially bad for ttjj). - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: CSC ttH(Hbb) further analysis

1

CSC ttH(Hbb) further analysis

Chris Collins-Tooth12/11/2008

Page 2: CSC ttH(Hbb) further analysis

2

Motivation

• Some late changes were made to the CSC note “cut-based” analysis.

• I felt that I didn’t have time to look at the effects on background shape (which were especially bad for ttjj).

• I also did not get chance to examine the effect of these cuts on truth-matching of reconstructed tLep, tHad, Higgs and WHad.

Page 3: CSC ttH(Hbb) further analysis

3

Effect of unwinding various late-stage cuts on distributions

• (a) CSC note distributions 30 fb-1…(”the problem”!)• (b) Relax b-tag cut from 5.5 to 4.5• (c) = (b)+consider ALL jets >4.5 as bJets (not just top four)• (d) =(c)+increase LightJet weights by +1.0 instead of +0.9• (e) =(d)+don’t rescale leptonic W before combining with a b-jet• (f) =(e)+Mt window from ±25GeV to ±20GeV• (g) =(f)+MWlep window from ±60GeV to ±25GeV

MH MH

(a)

Limited by ttjj statistics

Mean=118.5=22.85

Page 4: CSC ttH(Hbb) further analysis

4

Effect of unwinding various late-stage cuts on distributions

• (a) CSC note distributions 30 fb-1…(”the problem”!)• (b) Relax b-tag cut from 5.5 to 4.5• (c) = (b)+consider ALL jets >4.5 as bJets (not just top four)• (d) =(c)+increase LightJet weights by +1.0 instead of +0.9• (e) =(d)+don’t rescale leptonic W before combining with a b-jet• (f) =(e)+Mt window from ±25GeV to ±20GeV• (g) =(f)+MWlep window from ±60GeV to ±25GeV

Resolution of signal MH peak not reflected in this plot. Binning unhelpful too!

MH MH

(a)

Limited by ttjj statistics

Mean=118.5=22.85

Page 5: CSC ttH(Hbb) further analysis

5

Unwinding cuts..effect on ttjj

(a) ttjj CSC (b) btag4.5. (ttjj larger) (c) bthreshold off (d) LightJets wt+1.0

(e) Wlep rescale off (f) Mt ±20 GeV (g) MWlep ±25 GeV

Page 6: CSC ttH(Hbb) further analysis

6

Unwinding cuts..effect on ttjj

(a) ttjj CSC (b) btag4.5. (ttjj larger) (c) bthreshold off (d) LightJets wt+1.0

(e) Wlep rescale off (f) Mt ±20 GeV (g) MWlep ±25 GeV

Possibly the smoothest..Could be just binning/statistics Appears to peak lower than signal.. BUT 196/86 (2.3) times more ttjj

Slightly better S/B here

Page 7: CSC ttH(Hbb) further analysis

7

Unwinding cuts.. effect on signal

• Basically very little effect; – relaxing btag cut to 4.5

multiplies signal by 1.3 (c.f. 2.3 for ttjj)

• Other cuts give commensurate changes in Signal and Background,

– but give smoother background distributions.

• Slight exception: Mt-> ±20– appears to win a little S/B..

(S:22542033 ; B:196156)

– loses smoothness in ttjj shape terms (statistical?)

(a) CSC (b) btag4.5. (ttjj +larger) (c) bthreshold off (d) LightJets wt+1.0

(e) Wlep rescale off (f) Mt ±20 GeV (g) MWlep ±25 GeV

Page 8: CSC ttH(Hbb) further analysis

8

Distributions/Significances

• CSC Distributions• ..Ugly……• S=1.8• S/B=0.11

• Distribution (g) rebinned putting Mh=120 not at bin-edge

• Compromise..? =1.64• S/B=0.09

• Distribution (e)• Smoother…..• peaks all in same

place =1.61• S/B=0.07

Page 9: CSC ttH(Hbb) further analysis

9

Effect on truth-matching• For quarks having

– Pt> QuarkPtThreshold (2 GeV)– Matched to a jet within R<QuarkDeltaRCut (0.5)

• To be included, events had to have a reconstructed ttbar combination (inside mass windows).

Page 10: CSC ttH(Hbb) further analysis

10

Effect on truth-matching• For quarks having

– Pt> QuarkPtThreshold (2 GeV)– Matched to a jet within R<QuarkDeltaRCut (0.5)

• To be included, events had to have a reconstructed ttbar combination (inside mass windows).– NB: sometimes up to 6 light quarks come from WHad in the truth record;

• jet matching to these simply asks if the jet we used to reconstruct WHad was associated with a quark/quarks coming from Truth WHad..

• Reconstructed WHad only have 2 jet constituents.

Page 11: CSC ttH(Hbb) further analysis

11

Effect on truth-matching• For quarks having

– Pt> QuarkPtThreshold (2 GeV)– Matched to a jet within R<QuarkDeltaRCut (0.5)

• To be included, events had to have a reconstructed ttbar combination (inside mass windows).– NB: sometimes up to 6 light quarks come from WHad in the truth record;

• jet matching to these simply asks if the jet we used to reconstruct WHad was associated with a quark/quarks coming from Truth WHad..

• Reconstructed WHad only have 2 jet constituents.– NB: a reco jet used as (say) btLep might be associated with truth btLep; but

also might be associated with a truth btHad or bHiggs…or a quark from WHad. • To get over this: if the reconstructed item is associated to the truth item it’s

counted as ‘correct’.

Page 12: CSC ttH(Hbb) further analysis

12

Effect on truth-matching• For quarks having

– Pt> QuarkPtThreshold (2 GeV)– Matched to a jet within R<QuarkDeltaRCut (0.5)

• To be included, events had to have a reconstructed ttbar combination (inside mass windows).– NB: sometimes up to 6 light quarks come from WHad in the truth record;

• jet matching to these simply asks if the jet we used to reconstruct WHad was associated with a quark/quarks coming from Truth WHad..

• Reconstructed WHad only have 2 jet constituents.– NB: a reco jet used as (say) btLep might be associated with truth btLep; but

also might be associated with a truth btHad or bHiggs…or a quark from WHad. • To get over this: if the reconstructed item is associated to the truth item it’s

counted as ‘correct’.– Another problem (I have not solved!) is what to do when a jet is used in a

particle, but it does not match the truth record for that particle (happens often)..

• Then if there are two possible truth matches for that jet (say HiggsB and a LightQuark..)

• What do we say it ‘really’ came from? The closer of the two? This will skew the plots..at the moment it depends on the order of my ‘if’ statements!

Page 13: CSC ttH(Hbb) further analysis

13

Effect on truth-matching• For quarks having

– Pt> QuarkPtThreshold (2 GeV)– Matched to a jet within R<QuarkDeltaRCut (0.5)

• To be included, events had to have a reconstructed ttbar combination (inside mass windows).– NB: sometimes up to 6 light quarks come from WHad in the truth record;

• jet matching to these simply asks if the jet we used to reconstruct WHad was associated with a quark/quarks coming from Truth WHad..

• Reconstructed WHad only have 2 jet constituents.– NB: a reco jet used as (say) btLep might be associated with truth btLep; but

also might be associated with a truth btHad or bHiggs…or a quark from WHad. • To get over this: if the reconstructed item is associated to the truth item it’s

counted as ‘correct’.– Another problem (I have not solved!) is what to do when a jet is used in a

particle, but it does not match the truth record for that particle (happens often)..

• Then if there are two possible truth matches for that jet (say HiggsB and a LightQuark..)

• What do we say it ‘really’ came from? The closer of the two? This will skew the plots..at the moment it depends on the order of my ‘if’ statements!

• Uses?: Could look into events where we got combination right but didn’t make a Higgs; can we relax b-weight restriction for bb(H)???

Page 14: CSC ttH(Hbb) further analysis

14

CSC Signal truth assignments

• Large number of Higgs b-jets end up reconstructed in tLep or tHad• Worst reconstruction efficiency in building btHad • Slightly different to those produced by Georges Aad, but pretty

similar– Differences could be due to matching criteria– The results are close enough for me..

Jet used as btHad Jet used as bHiggsJet used as btLep Jet used as jWHad

tLep

tHad

H WH

ad

Unk

n

Jet came from

tLep

tHad

H WH

ad

Unk

n

tLep

tHad

H WH

ad

Unk

n

tLep tH

adH W

Had

Unk

n

Jet used as

Page 15: CSC ttH(Hbb) further analysis

15

Effect of changes (a)..(g).. small!(a) ttjj CSC

(b) btag4.5.

(c) bthreshold off

(d) LightJets wt+1.0

(e) Wlep rescale off

(f) Mt ±20 GeV

(g) MWlep ±25 GeV

Jet used as btLep Jet used as btHad Jet used as bHiggs Jet used as jWHad

tl thH WX tl thH WX tl thH WX tl thH WX

Page 16: CSC ttH(Hbb) further analysis

16

Effect of changes (a)..(g).. small!(a) ttjj CSC

(b) btag4.5.

(c) bthreshold off

(d) LightJets wt+1.0

(e) Wlep rescale off

(f) Mt ±20 GeV

(g) MWlep ±25 GeV

Jet used as btLep Jet used as btHad Jet used as bHiggs Jet used as jWHad

tl thH WX tl thH WX tl thH WX tl thH WX

Page 17: CSC ttH(Hbb) further analysis

17

Future studies• Future: would like to look at background truth-

assignments– Might be able to tell where the (fake) Higgs-b’s are coming from!

• Many further things I now intend to look at.. – Jet Algorithms (Kt4 would be easy to try.. Did some work a while

ago checking that b-tagging was OK to use for Kt..) – Relaxing Btag cut for Higgs b’s (after ttbar reconstructed)– Using Highest Pt top as opposed to best chi-squared top..

• might help with Higgs-b’s getting into tops

• Also look at further strategies for background measurement from data… – Getting smooth distributions is only one!