Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables

Shinei Seki, Yasushi Nagata

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    This study aims to improve the prediction accuracy for high-dimensional, small-sample-size data in a regression analysis. When using such data, scholars suggest the use of the cluster representative lasso that combines a cluster analysis and lasso, particularly when the covariance matrix has a block diagonal structure. In this study, we propose a new technique, called the graphical principal component lasso with focus on the block diagonal structure of the covariance matrix and latent variables. From the simulation results, we conclude that the proposed method is superior to the adaptive lasso, cluster representative lasso and principal component regression in terms of prediction accuracy for high-dimensional, small-sample-size data.

    Original languageEnglish
    Title of host publicationMathematical Methods and Computational Techniques in Science and Engineering II
    PublisherAmerican Institute of Physics Inc.
    Volume1982
    ISBN (Electronic)9780735416987
    DOIs
    Publication statusPublished - 2018 Jul 27
    Event2nd International Conference on Mathematical Methods and Computational Techniques in Science and Engineering - Cambridge, United Kingdom
    Duration: 2018 Feb 162018 Feb 18

    Other

    Other2nd International Conference on Mathematical Methods and Computational Techniques in Science and Engineering
    CountryUnited Kingdom
    CityCambridge
    Period18/2/1618/2/18

    Fingerprint

    data structures
    regression analysis
    cluster analysis
    predictions
    simulation

    ASJC Scopus subject areas

    • Physics and Astronomy(all)

    Cite this

    Seki, S., & Nagata, Y. (2018). Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables. In Mathematical Methods and Computational Techniques in Science and Engineering II (Vol. 1982). [020017] American Institute of Physics Inc.. https://doi.org/10.1063/1.5045423

    Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables. / Seki, Shinei; Nagata, Yasushi.

    Mathematical Methods and Computational Techniques in Science and Engineering II. Vol. 1982 American Institute of Physics Inc., 2018. 020017.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Seki, S & Nagata, Y 2018, Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables. in Mathematical Methods and Computational Techniques in Science and Engineering II. vol. 1982, 020017, American Institute of Physics Inc., 2nd International Conference on Mathematical Methods and Computational Techniques in Science and Engineering, Cambridge, United Kingdom, 18/2/16. https://doi.org/10.1063/1.5045423
    Seki S, Nagata Y. Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables. In Mathematical Methods and Computational Techniques in Science and Engineering II. Vol. 1982. American Institute of Physics Inc. 2018. 020017 https://doi.org/10.1063/1.5045423
    Seki, Shinei ; Nagata, Yasushi. / Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables. Mathematical Methods and Computational Techniques in Science and Engineering II. Vol. 1982 American Institute of Physics Inc., 2018.
    @inproceedings{c799fdc7b17d4018ad0de832d86a268a,
    title = "Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables",
    abstract = "This study aims to improve the prediction accuracy for high-dimensional, small-sample-size data in a regression analysis. When using such data, scholars suggest the use of the cluster representative lasso that combines a cluster analysis and lasso, particularly when the covariance matrix has a block diagonal structure. In this study, we propose a new technique, called the graphical principal component lasso with focus on the block diagonal structure of the covariance matrix and latent variables. From the simulation results, we conclude that the proposed method is superior to the adaptive lasso, cluster representative lasso and principal component regression in terms of prediction accuracy for high-dimensional, small-sample-size data.",
    author = "Shinei Seki and Yasushi Nagata",
    year = "2018",
    month = "7",
    day = "27",
    doi = "10.1063/1.5045423",
    language = "English",
    volume = "1982",
    booktitle = "Mathematical Methods and Computational Techniques in Science and Engineering II",
    publisher = "American Institute of Physics Inc.",

    }

    TY - GEN

    T1 - Regression analysis on high-dimensional, block diagonal structure data with focus on latent variables

    AU - Seki, Shinei

    AU - Nagata, Yasushi

    PY - 2018/7/27

    Y1 - 2018/7/27

    N2 - This study aims to improve the prediction accuracy for high-dimensional, small-sample-size data in a regression analysis. When using such data, scholars suggest the use of the cluster representative lasso that combines a cluster analysis and lasso, particularly when the covariance matrix has a block diagonal structure. In this study, we propose a new technique, called the graphical principal component lasso with focus on the block diagonal structure of the covariance matrix and latent variables. From the simulation results, we conclude that the proposed method is superior to the adaptive lasso, cluster representative lasso and principal component regression in terms of prediction accuracy for high-dimensional, small-sample-size data.

    AB - This study aims to improve the prediction accuracy for high-dimensional, small-sample-size data in a regression analysis. When using such data, scholars suggest the use of the cluster representative lasso that combines a cluster analysis and lasso, particularly when the covariance matrix has a block diagonal structure. In this study, we propose a new technique, called the graphical principal component lasso with focus on the block diagonal structure of the covariance matrix and latent variables. From the simulation results, we conclude that the proposed method is superior to the adaptive lasso, cluster representative lasso and principal component regression in terms of prediction accuracy for high-dimensional, small-sample-size data.

    UR - http://www.scopus.com/inward/record.url?scp=85051119866&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85051119866&partnerID=8YFLogxK

    U2 - 10.1063/1.5045423

    DO - 10.1063/1.5045423

    M3 - Conference contribution

    VL - 1982

    BT - Mathematical Methods and Computational Techniques in Science and Engineering II

    PB - American Institute of Physics Inc.

    ER -