Bayesian inference from case-cohort data with multiple end-points

    Research output: Contribution to journalArticleScientificpeer-review

    Abstract

    "In a case-cohort design a random sample from the study cohort, referred as a subcohort, and all the cases outside the subcohort are selected for collecting extra covariate data. The union of the selected subcohort and all cases are referred as the case-cohort set. Such a design is generally employed when the collection of information on an extra covariate for the study cohort is expensive. An advantage of the case-cohort design over more traditional case-control and the nested case-control designs is that it provides a set of controls which can be used for multiple end-points, in which case there is information on some covariates; and event follow-up for the whole study cohort. Here, we propose a Bayesian approach to analyse such a case-cohort design as a cohort design with incomplete data on the extra covariate. We construct likelihood expressions when multiple endpoints are of interest simultaneously and propose a Bayesian data augmentation method to estimate the model parameters. A simulation study is carried out to illustrate the method and the results are compared with the complete cohort analysis."
    Original languageEnglish
    JournalScandinavian Journal of Statistics
    Volume33
    Issue number1
    Pages (from-to)25-36
    Number of pages12
    ISSN0303-6898
    DOIs
    Publication statusPublished - 2006
    MoE publication typeA1 Journal article-refereed

    Fields of Science

    • 111 Mathematics

    Cite this

    @article{aa32e020be444d40ab3f4a27692643ed,
    title = "Bayesian inference from case-cohort data with multiple end-points",
    abstract = "{"}In a case-cohort design a random sample from the study cohort, referred as a subcohort, and all the cases outside the subcohort are selected for collecting extra covariate data. The union of the selected subcohort and all cases are referred as the case-cohort set. Such a design is generally employed when the collection of information on an extra covariate for the study cohort is expensive. An advantage of the case-cohort design over more traditional case-control and the nested case-control designs is that it provides a set of controls which can be used for multiple end-points, in which case there is information on some covariates; and event follow-up for the whole study cohort. Here, we propose a Bayesian approach to analyse such a case-cohort design as a cohort design with incomplete data on the extra covariate. We construct likelihood expressions when multiple endpoints are of interest simultaneously and propose a Bayesian data augmentation method to estimate the model parameters. A simulation study is carried out to illustrate the method and the results are compared with the complete cohort analysis.{"}",
    keywords = "111 Mathematics",
    author = "Sangita Kulathinal and Elja Arjas",
    year = "2006",
    doi = "10.1111/j.1467-9469.2006.00481.x",
    language = "English",
    volume = "33",
    pages = "25--36",
    journal = "Scandinavian Journal of Statistics",
    issn = "0303-6898",
    publisher = "Wiley",
    number = "1",

    }

    Bayesian inference from case-cohort data with multiple end-points. / Kulathinal, Sangita; Arjas, Elja.

    In: Scandinavian Journal of Statistics, Vol. 33, No. 1, 2006, p. 25-36.

    Research output: Contribution to journalArticleScientificpeer-review

    TY - JOUR

    T1 - Bayesian inference from case-cohort data with multiple end-points

    AU - Kulathinal, Sangita

    AU - Arjas, Elja

    PY - 2006

    Y1 - 2006

    N2 - "In a case-cohort design a random sample from the study cohort, referred as a subcohort, and all the cases outside the subcohort are selected for collecting extra covariate data. The union of the selected subcohort and all cases are referred as the case-cohort set. Such a design is generally employed when the collection of information on an extra covariate for the study cohort is expensive. An advantage of the case-cohort design over more traditional case-control and the nested case-control designs is that it provides a set of controls which can be used for multiple end-points, in which case there is information on some covariates; and event follow-up for the whole study cohort. Here, we propose a Bayesian approach to analyse such a case-cohort design as a cohort design with incomplete data on the extra covariate. We construct likelihood expressions when multiple endpoints are of interest simultaneously and propose a Bayesian data augmentation method to estimate the model parameters. A simulation study is carried out to illustrate the method and the results are compared with the complete cohort analysis."

    AB - "In a case-cohort design a random sample from the study cohort, referred as a subcohort, and all the cases outside the subcohort are selected for collecting extra covariate data. The union of the selected subcohort and all cases are referred as the case-cohort set. Such a design is generally employed when the collection of information on an extra covariate for the study cohort is expensive. An advantage of the case-cohort design over more traditional case-control and the nested case-control designs is that it provides a set of controls which can be used for multiple end-points, in which case there is information on some covariates; and event follow-up for the whole study cohort. Here, we propose a Bayesian approach to analyse such a case-cohort design as a cohort design with incomplete data on the extra covariate. We construct likelihood expressions when multiple endpoints are of interest simultaneously and propose a Bayesian data augmentation method to estimate the model parameters. A simulation study is carried out to illustrate the method and the results are compared with the complete cohort analysis."

    KW - 111 Mathematics

    U2 - 10.1111/j.1467-9469.2006.00481.x

    DO - 10.1111/j.1467-9469.2006.00481.x

    M3 - Article

    VL - 33

    SP - 25

    EP - 36

    JO - Scandinavian Journal of Statistics

    JF - Scandinavian Journal of Statistics

    SN - 0303-6898

    IS - 1

    ER -