1. Theoretical Motivation Analysis Procedure Systematics Results
David Doll,
on behalf of the
BaBar Collaboration
APS 04/12/08

2. Theoretical Motivation
Standard Model BF
Experimental Limit on BF (90% CL)*
arXiv: v2 [hep-ex],
PRL 99, (2007)
*535 M pairs at Belle
Highly suppressed Flavor Changing Neutral Current
Not well constrained experimentally
Several models enhance BF(Unparticle Model, MSSM at large tan β,…)
BaBar’s previous best upper limit is 7.8x10-5 for semileptonic tags with 81.9 fb-1
Current analysis at 319 fb-1
APS 04/12/08

3. Analysis Procedure, Tagging
Perform a ‘semileptonic’ tagged analysis
Fully reconstruct the ‘tag B’ in the decay
Look at the rest of the event for our signal B+ B-
Tag B
Signal B
APS 04/12/08

4. Random Forest (RF)
Use a multivariate analysis tool from StatPatternRecognition (arXiv:physics/ v1)
Sampling with replacement of both the training data and the input variables (bagging)
Optimize the ‘Punzi’ Figure of Merit
The important input variables:
number of charged tracks in the signal B (opposite the ‘tag B’)
the missing energy in the event
the signal Kaon candidate’s momentum
the unmatched neutral energy in the event
APS B->Knunu 04/12/08

5. Final Predictions, Continuum
RF continuum est.
Use sideband region in
Estimate the continuum data in the signal region from amount of data in sideband
sideband
APS B->Knunu 04/12/08

6. Final Predictions, Peaking
Peaking estimate from RF output, separated into sideband/signal regions
Subtract sideband from signal region in both Data and MC and take the ratio MC:Data
Extrapolate a line into the signal region
signal region
trendline
APS B->Knunu 04/12/08

7. Background Systematics
Continuum systematic from difference between MC and data
Peaking background systematic from difference between the a trendline fit to all the MC:Data, vs. a trendline fit to just the peaking component (above)
We also take a systematic based on our MC weighting procedure.
MC Background prediction
Statistical Uncertainty
Systematic uncertainty
APS B->Knunu 04/12/08

8. Control Sample B- B+ Both Bs decay semileptonicly requiring:
no remaining charged tracks in the event
momentum of each lepton>1.24 GeV/c
Resolved differences between signal MC and double tag data:
particle substitutions
kinematic corrections
brute force variable redistribution.
Serves as control sample for evaluating systematics for the multivariate analysis. B+ B-
APS B->Knunu 04/12/08

9. Signal Systematics
Tagging Efficiency: Taken from ratio below in which both tags are
Kaon Momentum: Evaluated by comparing phase space theory with SM-predicted theory
APS B->Knunu 04/12/08

10. Signal Systematics Correlations btwn. Variables:
1-D distributions already resolved
Need to account for correlations in order use the control sample to evaluate signal box efficiency in signal MC
APS B->Knunu 04/12/08

11. Signal Systematics Signal Box Eff.: Ntrkleft=1:
Retrain RF with double tag MC control sample substituted for signal MC
Evaluate systematic by comparing efficiency of the RF cut on double tag MC to double tag data
Ntrkleft=1:
The control sample identified with this cut, not present in signal MC
Evaluate systematic from separate rectangular cut based investigation
APS B->Knunu 04/12/08

12. Results Upper limit at the 90% confidence level
Expect / events, corresponding to an upper limit of 2.9 x 10-5
Inside the RF box, we saw 38 events, which gives an upper limit:
APS B->Knunu 04/12/08