some aspects of statistics at lhc
DESCRIPTION
Some aspects of statistics at LHC. N.V.Krasnikov INR, Moscow. Outline. 1. Introduction 2. Parameters estimation 3. Confidence intervals 4. Systematics 5. Upper limits at LHC 6. Conclusion. 1. Introduction The statistical model of an analysis - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/1.jpg)
October 2013
Some aspects of statistics at LHC
N.V.Krasnikov
INR, Moscow
![Page 2: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/2.jpg)
October 2013
1. Introduction
2. Parameters estimation
3. Confidence intervals
4. Systematics
5. Upper limits at LHC
6. Conclusion
Outline
![Page 3: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/3.jpg)
October 2013
1. Introduction
The statistical model of an analysisprovides the complete description of thatanalysisThe main problem – from the known probability density and x = xobs to extract some information
on θ parameterTwo approaches 1.Frequentist method2. Bayesian methodAlso very important – the notion of the likelihood
![Page 4: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/4.jpg)
October 2013
Likelihood - the probability density evaluated at the observed value x=xobs
![Page 5: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/5.jpg)
Frequents statistics – general philosophy
In frequentist statistics, probabilities are associated only with data, i.e. outcomes of
repeatable observations. The preferred models are those for which our observations
have non small probabilities
October 2013
![Page 6: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/6.jpg)
October 2013
![Page 7: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/7.jpg)
Bayesian statistics – general philosophy
In Bayesian statistics , interpretation of probability is extended to degree of
belief(subjective probability). Bayesian methods can provide more natural treatment
of non repeatable phenomena :
systematic uncertainties
October 2013
![Page 8: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/8.jpg)
October 2013
Parameters estimation
Maximum likelihood principle
![Page 9: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/9.jpg)
October 2013
Normal distribution
![Page 10: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/10.jpg)
October 2013
Bayesian method
• In Bayes approach
For flat prior π(θ) = const
Bayes and likelihood coincide
![Page 11: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/11.jpg)
October 2013
Confidence intervals
Suppose we measure x = xobs
• What are possible values of θ parameter?
• Frequentist answer:
Neyman belt construction
Alternative:
Bayes credible interval
![Page 12: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/12.jpg)
October 2013
Neyman belt construction
(1-α) – confidence level. The choice of x1 and x2
is not unique
![Page 13: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/13.jpg)
Neyman belt construction
October 2013
![Page 14: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/14.jpg)
October 2013
Neyman belt construction
![Page 15: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/15.jpg)
October 2013
Neyman belt construction
• For normal distribution Neyman belt equations for lower limit lead to
![Page 16: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/16.jpg)
October 2013
Neyman belt equations
![Page 17: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/17.jpg)
October 2013
Maximal likelihood Approximate estimate
For normal distribution
![Page 18: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/18.jpg)
October 2013
Bayes approach
Bayes theorem
P(A|B)*P(B) = P(B|A)*P(B)
P(A|B) – conditional probability
![Page 19: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/19.jpg)
October 2013
Bayes approach• Due to Bayes formula
the statistics problem is reduced to the
probability problem
![Page 20: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/20.jpg)
October 2013
Bayes approach• The main problem – prior function π(θ) is
not known • For what prior frequentist and Bayes approaches
coincide?
![Page 21: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/21.jpg)
October 2013
The relation between Bayes and frequentist approaches
• Two examples
1.Example A
![Page 22: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/22.jpg)
October 2013
The relation between Bayes and frequentist approaches
2.Example B
![Page 23: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/23.jpg)
October 2013
Parameter determination with additional constraint
• Consider the case of normal distribution
with additional constraint
Maximum likelihood method gives
![Page 24: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/24.jpg)
October 2013
Parameter determination with additional constraint
• How to construct the confidence interval for the μ parameter?
Cousins, Feldman method• Maximum of
![Page 25: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/25.jpg)
October 2013
Neyman belt construction
• The ordering principle on • As a consequence we find
![Page 26: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/26.jpg)
October 2013
Likelihood method
• For x0 < 0
• or
![Page 27: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/27.jpg)
October 2013
Likelihood principle
• For x0 > 0
• or
![Page 28: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/28.jpg)
October 2013
Bayes approach
• We choose π(μ) = θ(μ)
So prior function is zero for negative μ
automatically• The equation for the credible interval
determination
![Page 29: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/29.jpg)
October 2013
Confidence intervals for Poisson distribution
• The generalization of Neyman belt construction is
Klopper-Pearson interval
![Page 30: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/30.jpg)
October 2013
Poisson distribution
• The Kloper-Pearson interval is conservative and it does not have the coverage property. Coverage is the probability that interval covers true value with the probability Besides for
So we have negative probability - contradiction
![Page 31: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/31.jpg)
October 2013
Stevens interval
• To overcome these problems Stevens (1952) suggested to introduce new random variable U. Modified equations are
![Page 32: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/32.jpg)
October 2013
Stevens equations• One can derive Stevens equations using the
regularization of discrete Poisson distribution(S.Bityukov,N.V.K). Namely let us introduce Poisson generalization
• The integral
• is not well defined
![Page 33: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/33.jpg)
October 2013
Stevens interval
Let us introduce the regularization
![Page 34: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/34.jpg)
October 2013
Stevens interval
• We can use Neyman belt construction for regularized distribution and we find
![Page 35: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/35.jpg)
Stevens intervalIn the limit of the regularization removement
we find
October 2013
![Page 36: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/36.jpg)
October 2013
Likelihood method
• The use of likelihood method gives
• The solution is
![Page 37: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/37.jpg)
October 2013
Likelihood method
![Page 38: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/38.jpg)
October 2013
Bayes approach
• The basic equations are
• Due to identities
![Page 39: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/39.jpg)
October 2013
Bayes approach
Upper Klopper-Pearson limit coincides
with Bayesian limit for flat prior and lower limit corresponds to prior
The Stevens equations for non
dependent on λ
are equivalent to Bayes approach with prior
function
![Page 40: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/40.jpg)
October 2013
Uncertainties in extraction of an upper limit
![Page 41: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/41.jpg)
October 2013
Modified frequentist definition• We require(S.Bityukov,N.V.K.,2012) that
• Our definition is equivalent to Bayes • approach with prior function
![Page 42: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/42.jpg)
October 2013
Signal extraction for nonzero background
• Consider the case
• Cousins-Feldman method
![Page 43: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/43.jpg)
October 2013
Nonzero background
• Likelihood ordering
Plus Neyman construction
![Page 44: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/44.jpg)
October 2013
CLS method(T.Junk,A.Read)
• Upper bound
• CLS method
• In Bayes approach it corresponds to the replacement
![Page 45: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/45.jpg)
October 2013
Bayes method
• For flat prior
• We can interprete this formula in terms of conditional probability
![Page 46: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/46.jpg)
October 2013
Bayes method
• Namely the probability that parameter λ lies in the interval [λ, λ+dλ] provided λ≥λb is determined by the formula
that coincides with the previous Bayes formula
![Page 47: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/47.jpg)
October 2013
Systematics
• 1. Systematics that can be eliminated by the measurement of some variables in other kinematic region
• 2. Uncertainties related with nonexact accuracy in determination of particle momenta, misidentification...
• 3. Uncertainties related with nonexact knowledge of theoretical cross sections
![Page 48: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/48.jpg)
October 2013
Systematics• 3 methods to deal with systematics(at least)
1. Suppose we measure some events in
two kinematic regions with distribution
functions + ,
The random variable Z = X-Y obeys normal distribution As a consequence we find
![Page 49: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/49.jpg)
October 2013
Systematics
• For Poisson distributions
and
due to identity
![Page 50: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/50.jpg)
October 2013
Systematics
• The problem is reduced to the determination of the ρ parameter from experimental data
![Page 51: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/51.jpg)
October 2013
Systematics
2.Bayesian treatment or Cousins-Highland
method is based on integration over nonessential variables
For normal distributions and flat prior we find
![Page 52: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/52.jpg)
October 2013
Systematics
• In other words the main effect is the replacement
and the significance is
So for normal distribution this method coincides with the first method
![Page 53: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/53.jpg)
October 2013
Systematics
• Profile likelihood method
Suppose likelihood function
depends on nonessential variables θ
and essential variables λ
Profile likelihood
![Page 54: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/54.jpg)
October 2013
Profile likelihood
New variable(statistics)
• Per construction • For new statistics defines probability density
![Page 55: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/55.jpg)
October 2013
Profile likelihood
• For normal distributions profile likelihood method coincides with the Cousins-Highland method
• Very often p-value is used• By definition
p-value determines the agreement of data with amodel• Small p-value(p < 5.9*10-7) - the model is
excluded by experimental data
![Page 56: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/56.jpg)
October 2013
P-value
• For Poisson distribution p-value definition is
![Page 57: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/57.jpg)
October 2013
Limits on new physics at LHC
For the Higgs boson search CMS and ATLAS introduce the extended model
with additional μ parameter and the replacement cross section the same. The case μ =1 corresponds to SM. The case μ=0 corresponds to the absence of the SM Higgs boson.
The likelihood of the general model can be written in the form
![Page 58: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/58.jpg)
October 2013
Likelihood of the model
Here is the probability density of nonessential parameters . Usually
Is taken as normal or lognormal distribution
![Page 59: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/59.jpg)
October 2013
Bayes approach
• In Bayes approach the use of the formula
• allows to determine the probability density for μ parameter. Upper limit μup is detemined as
Usually α= 0.05
![Page 60: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/60.jpg)
October 2013
Frequentist approach
• CMS and ATLAS use statistics
Often modifications are used with additional conditions as
![Page 61: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/61.jpg)
October 2013
Frequentist approach
Very often the hypothesis μ=0 is tested
against μ>0. For such case it is convenient to use
For single Poisson
![Page 62: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/62.jpg)
October 2013
Single Poisson
• By construction q0≥0 and
In the limit nobs»1 the
probability density is
![Page 63: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/63.jpg)
October 2013
Upper limits
• To derive upper limits the statistics
is used. For single Poisson
![Page 64: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/64.jpg)
Higgs boson search at CMS
As an illustration consider the Higgs boson search at CMS detector
October 2013
![Page 65: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/65.jpg)
October 2013
P-value for Higgs boson search
![Page 66: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/66.jpg)
October 2013
![Page 67: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/67.jpg)
Summary of Higgs boson measurements
October 2013
![Page 68: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/68.jpg)
October 2013
![Page 69: Some aspects of statistics at LHC](https://reader036.vdocument.in/reader036/viewer/2022062304/56813b72550346895da47bb0/html5/thumbnails/69.jpg)
Conclusions
Experiments CMS and ATLAS use both frequentist and Bayesian
methods to extract the parameters of Higgs boson and limits on new
physics. As a rule they give numerically similar results
October 2013