Generalized information criterion

Masanobu Taniguchi, Junichi Hirukawa

    Research output: Contribution to journalArticle

    1 Citation (Scopus)

    Abstract

    In this article, we propose a generalized Akaike's information criterion (AIC) (GAIC), which includes the usual AIC as a special case, for general class of stochastic models (i.e. i.i.d., non-i.i.d., time series models etc.). Then we derive the asymptotic distribution of selected order by GAIC, and show that is inconsistent, i.e. (true order). This is the problem of selection by completely specified models. In practice, it is natural to suppose that the true model g would be incompletely specified by uncertain prior information, and be contiguous to a fundamental parametric model with dimθ 0=p 0. One plausible parametric description for g is , h=(h 1,...,h K-p 0)′ where n is the sample size, and the true order is K. Under this setting, we derive the asymptotic distribution of Then it is shown that GAIC has admissible properties for perturbation of models with order of , where the length {norm of matrix}h{norm of matrix} is large. This observation seems important. Also numerical studies will be given to confirm the results.

    Original languageEnglish
    Pages (from-to)287-297
    Number of pages11
    JournalJournal of Time Series Analysis
    Volume33
    Issue number2
    DOIs
    Publication statusPublished - 2012 Mar

    Fingerprint

    Information Criterion
    Akaike Information Criterion
    Asymptotic distribution
    Norm
    Time Series Models
    Prior Information
    Parametric Model
    Inconsistent
    Stochastic Model
    Numerical Study
    Stochastic models
    Sample Size
    Model
    Time series
    Perturbation
    Information criterion
    Akaike information criterion

    Keywords

    • AIC
    • Asymptotic theory
    • Information criterion
    • Model selection
    • Spectral distribution

    ASJC Scopus subject areas

    • Applied Mathematics
    • Statistics and Probability
    • Statistics, Probability and Uncertainty

    Cite this

    Generalized information criterion. / Taniguchi, Masanobu; Hirukawa, Junichi.

    In: Journal of Time Series Analysis, Vol. 33, No. 2, 03.2012, p. 287-297.

    Research output: Contribution to journalArticle

    Taniguchi, Masanobu ; Hirukawa, Junichi. / Generalized information criterion. In: Journal of Time Series Analysis. 2012 ; Vol. 33, No. 2. pp. 287-297.
    @article{29ffbebef9c74781a0a18b2fa6dc2911,
    title = "Generalized information criterion",
    abstract = "In this article, we propose a generalized Akaike's information criterion (AIC) (GAIC), which includes the usual AIC as a special case, for general class of stochastic models (i.e. i.i.d., non-i.i.d., time series models etc.). Then we derive the asymptotic distribution of selected order by GAIC, and show that is inconsistent, i.e. (true order). This is the problem of selection by completely specified models. In practice, it is natural to suppose that the true model g would be incompletely specified by uncertain prior information, and be contiguous to a fundamental parametric model with dimθ 0=p 0. One plausible parametric description for g is , h=(h 1,...,h K-p 0)′ where n is the sample size, and the true order is K. Under this setting, we derive the asymptotic distribution of Then it is shown that GAIC has admissible properties for perturbation of models with order of , where the length {norm of matrix}h{norm of matrix} is large. This observation seems important. Also numerical studies will be given to confirm the results.",
    keywords = "AIC, Asymptotic theory, Information criterion, Model selection, Spectral distribution",
    author = "Masanobu Taniguchi and Junichi Hirukawa",
    year = "2012",
    month = "3",
    doi = "10.1111/j.1467-9892.2011.00759.x",
    language = "English",
    volume = "33",
    pages = "287--297",
    journal = "Journal of Time Series Analysis",
    issn = "0143-9782",
    publisher = "Wiley-Blackwell",
    number = "2",

    }

    TY - JOUR

    T1 - Generalized information criterion

    AU - Taniguchi, Masanobu

    AU - Hirukawa, Junichi

    PY - 2012/3

    Y1 - 2012/3

    N2 - In this article, we propose a generalized Akaike's information criterion (AIC) (GAIC), which includes the usual AIC as a special case, for general class of stochastic models (i.e. i.i.d., non-i.i.d., time series models etc.). Then we derive the asymptotic distribution of selected order by GAIC, and show that is inconsistent, i.e. (true order). This is the problem of selection by completely specified models. In practice, it is natural to suppose that the true model g would be incompletely specified by uncertain prior information, and be contiguous to a fundamental parametric model with dimθ 0=p 0. One plausible parametric description for g is , h=(h 1,...,h K-p 0)′ where n is the sample size, and the true order is K. Under this setting, we derive the asymptotic distribution of Then it is shown that GAIC has admissible properties for perturbation of models with order of , where the length {norm of matrix}h{norm of matrix} is large. This observation seems important. Also numerical studies will be given to confirm the results.

    AB - In this article, we propose a generalized Akaike's information criterion (AIC) (GAIC), which includes the usual AIC as a special case, for general class of stochastic models (i.e. i.i.d., non-i.i.d., time series models etc.). Then we derive the asymptotic distribution of selected order by GAIC, and show that is inconsistent, i.e. (true order). This is the problem of selection by completely specified models. In practice, it is natural to suppose that the true model g would be incompletely specified by uncertain prior information, and be contiguous to a fundamental parametric model with dimθ 0=p 0. One plausible parametric description for g is , h=(h 1,...,h K-p 0)′ where n is the sample size, and the true order is K. Under this setting, we derive the asymptotic distribution of Then it is shown that GAIC has admissible properties for perturbation of models with order of , where the length {norm of matrix}h{norm of matrix} is large. This observation seems important. Also numerical studies will be given to confirm the results.

    KW - AIC

    KW - Asymptotic theory

    KW - Information criterion

    KW - Model selection

    KW - Spectral distribution

    UR - http://www.scopus.com/inward/record.url?scp=84858002747&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84858002747&partnerID=8YFLogxK

    U2 - 10.1111/j.1467-9892.2011.00759.x

    DO - 10.1111/j.1467-9892.2011.00759.x

    M3 - Article

    VL - 33

    SP - 287

    EP - 297

    JO - Journal of Time Series Analysis

    JF - Journal of Time Series Analysis

    SN - 0143-9782

    IS - 2

    ER -