Software for statistical data analysis used in Higgs searches

Christian Gumpert, Lorenzo Moneta, Kyle Cranmer, Sven Kreiss, Wouter Verkerke

    Research output: Contribution to journalArticle

    Abstract

    The analysis and interpretation of data collected by the Large Hadron Collider (LHC) requires advanced statistical tools in order to quantify the agreement between observation and theoretical models. RooStats is a project providing a statistical framework for data analysis with the focus on discoveries, confidence intervals and combination of different measurements in both Bayesian and frequentist approaches. It employs the RooFit data modelling language where mathematical concepts such as variables, (probability density) functions and integrals are represented as C++ objects. RooStats and RooFit rely on the persistency technology of the ROOT framework. The usage of a common data format enables the concept of digital publishing of complicated likelihood functions. The statistical tools have been developed in close collaboration with the LHC experiments to ensure their applicability to real-life use cases. Numerous physics results have been produced using the RooStats tools, with the discovery of the Higgs boson by the ATLAS and CMS experiments being certainly the most popular among them. We will discuss tools currently used by LHC experiments to set exclusion limits, to derive confidence intervals and to estimate discovery significances based on frequentist statistics and the asymptotic behaviour of likelihood functions. Furthermore, new developments in RooStats and performance optimisation necessary to cope with complex models depending on more than 1000 variables will be reviewed.

    Original languageEnglish (US)
    Article number012229
    JournalJournal of Physics: Conference Series
    Volume490
    Issue number1
    DOIs
    StatePublished - 2014

    Fingerprint

    computer programs
    confidence
    intervals
    probability density functions
    exclusion
    Higgs bosons
    format
    statistics
    physics
    optimization
    estimates

    ASJC Scopus subject areas

    • Physics and Astronomy(all)

    Cite this

    Software for statistical data analysis used in Higgs searches. / Gumpert, Christian; Moneta, Lorenzo; Cranmer, Kyle; Kreiss, Sven; Verkerke, Wouter.

    In: Journal of Physics: Conference Series, Vol. 490, No. 1, 012229, 2014.

    Research output: Contribution to journalArticle

    Gumpert, C, Moneta, L, Cranmer, K, Kreiss, S & Verkerke, W 2014, 'Software for statistical data analysis used in Higgs searches', Journal of Physics: Conference Series, vol. 490, no. 1, 012229. https://doi.org/10.1088/1742-6596/490/1/012229
    Gumpert, Christian ; Moneta, Lorenzo ; Cranmer, Kyle ; Kreiss, Sven ; Verkerke, Wouter. / Software for statistical data analysis used in Higgs searches. In: Journal of Physics: Conference Series. 2014 ; Vol. 490, No. 1.
    @article{6f1632ee4c5a4a479b29af5e8a0bf564,
    title = "Software for statistical data analysis used in Higgs searches",
    abstract = "The analysis and interpretation of data collected by the Large Hadron Collider (LHC) requires advanced statistical tools in order to quantify the agreement between observation and theoretical models. RooStats is a project providing a statistical framework for data analysis with the focus on discoveries, confidence intervals and combination of different measurements in both Bayesian and frequentist approaches. It employs the RooFit data modelling language where mathematical concepts such as variables, (probability density) functions and integrals are represented as C++ objects. RooStats and RooFit rely on the persistency technology of the ROOT framework. The usage of a common data format enables the concept of digital publishing of complicated likelihood functions. The statistical tools have been developed in close collaboration with the LHC experiments to ensure their applicability to real-life use cases. Numerous physics results have been produced using the RooStats tools, with the discovery of the Higgs boson by the ATLAS and CMS experiments being certainly the most popular among them. We will discuss tools currently used by LHC experiments to set exclusion limits, to derive confidence intervals and to estimate discovery significances based on frequentist statistics and the asymptotic behaviour of likelihood functions. Furthermore, new developments in RooStats and performance optimisation necessary to cope with complex models depending on more than 1000 variables will be reviewed.",
    author = "Christian Gumpert and Lorenzo Moneta and Kyle Cranmer and Sven Kreiss and Wouter Verkerke",
    year = "2014",
    doi = "10.1088/1742-6596/490/1/012229",
    language = "English (US)",
    volume = "490",
    journal = "Journal of Physics: Conference Series",
    issn = "1742-6588",
    publisher = "IOP Publishing Ltd.",
    number = "1",

    }

    TY - JOUR

    T1 - Software for statistical data analysis used in Higgs searches

    AU - Gumpert, Christian

    AU - Moneta, Lorenzo

    AU - Cranmer, Kyle

    AU - Kreiss, Sven

    AU - Verkerke, Wouter

    PY - 2014

    Y1 - 2014

    N2 - The analysis and interpretation of data collected by the Large Hadron Collider (LHC) requires advanced statistical tools in order to quantify the agreement between observation and theoretical models. RooStats is a project providing a statistical framework for data analysis with the focus on discoveries, confidence intervals and combination of different measurements in both Bayesian and frequentist approaches. It employs the RooFit data modelling language where mathematical concepts such as variables, (probability density) functions and integrals are represented as C++ objects. RooStats and RooFit rely on the persistency technology of the ROOT framework. The usage of a common data format enables the concept of digital publishing of complicated likelihood functions. The statistical tools have been developed in close collaboration with the LHC experiments to ensure their applicability to real-life use cases. Numerous physics results have been produced using the RooStats tools, with the discovery of the Higgs boson by the ATLAS and CMS experiments being certainly the most popular among them. We will discuss tools currently used by LHC experiments to set exclusion limits, to derive confidence intervals and to estimate discovery significances based on frequentist statistics and the asymptotic behaviour of likelihood functions. Furthermore, new developments in RooStats and performance optimisation necessary to cope with complex models depending on more than 1000 variables will be reviewed.

    AB - The analysis and interpretation of data collected by the Large Hadron Collider (LHC) requires advanced statistical tools in order to quantify the agreement between observation and theoretical models. RooStats is a project providing a statistical framework for data analysis with the focus on discoveries, confidence intervals and combination of different measurements in both Bayesian and frequentist approaches. It employs the RooFit data modelling language where mathematical concepts such as variables, (probability density) functions and integrals are represented as C++ objects. RooStats and RooFit rely on the persistency technology of the ROOT framework. The usage of a common data format enables the concept of digital publishing of complicated likelihood functions. The statistical tools have been developed in close collaboration with the LHC experiments to ensure their applicability to real-life use cases. Numerous physics results have been produced using the RooStats tools, with the discovery of the Higgs boson by the ATLAS and CMS experiments being certainly the most popular among them. We will discuss tools currently used by LHC experiments to set exclusion limits, to derive confidence intervals and to estimate discovery significances based on frequentist statistics and the asymptotic behaviour of likelihood functions. Furthermore, new developments in RooStats and performance optimisation necessary to cope with complex models depending on more than 1000 variables will be reviewed.

    UR - http://www.scopus.com/inward/record.url?scp=84896947162&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84896947162&partnerID=8YFLogxK

    U2 - 10.1088/1742-6596/490/1/012229

    DO - 10.1088/1742-6596/490/1/012229

    M3 - Article

    VL - 490

    JO - Journal of Physics: Conference Series

    JF - Journal of Physics: Conference Series

    SN - 1742-6588

    IS - 1

    M1 - 012229

    ER -