Prediction error in partial least squares (PLS) regression: a critique on the deviation used in The Unscrambler

S. de Vries, C.J.F. ter Braak

    Research output: Contribution to journalArticleAcademicpeer-review

    50 Citations (Scopus)

    Abstract

    Partial least squares (PLS) regression is commonly used for multivariate calibration of instruments. Because of the need to know the quality of the prediction in a specific unknown sample and the lack of theory, an ‘empirically found formula’ to express the uncertainty is utilized in The Unscrambler II software, the de-facto standard in computer software for PLS. In this critique the formula is examined theoretically and by simulation. It is concluded that this formula underestimates the root mean squared error of prediction in most practical applications of PLS. A change of the formula is planned in the next version of The Unscrambler. In the mean time users of The Unscrambler ver 5.5 or lower should multiply the reported deviation by a factor of at least , to get a reasonable estimate of the prediction error
    Original languageEnglish
    Pages (from-to)239-245
    JournalChemometrics and Intelligent Laboratory Systems
    Volume30
    DOIs
    Publication statusPublished - 1995

    Fingerprint

    Calibration
    Uncertainty

    Keywords

    • PLS
    • The Unscrambler

    Cite this

    @article{7e29438bfdb8465f98b71965edaa2be3,
    title = "Prediction error in partial least squares (PLS) regression: a critique on the deviation used in The Unscrambler",
    abstract = "Partial least squares (PLS) regression is commonly used for multivariate calibration of instruments. Because of the need to know the quality of the prediction in a specific unknown sample and the lack of theory, an ‘empirically found formula’ to express the uncertainty is utilized in The Unscrambler II software, the de-facto standard in computer software for PLS. In this critique the formula is examined theoretically and by simulation. It is concluded that this formula underestimates the root mean squared error of prediction in most practical applications of PLS. A change of the formula is planned in the next version of The Unscrambler. In the mean time users of The Unscrambler ver 5.5 or lower should multiply the reported deviation by a factor of at least , to get a reasonable estimate of the prediction error",
    keywords = "PLS, The Unscrambler, PLS",
    author = "{de Vries}, S. and {ter Braak}, C.J.F.",
    year = "1995",
    doi = "10.1016/0169-7439(95)00030-5",
    language = "English",
    volume = "30",
    pages = "239--245",
    journal = "Chemometrics and Intelligent Laboratory Systems",
    issn = "0169-7439",
    publisher = "Elsevier",

    }

    Prediction error in partial least squares (PLS) regression: a critique on the deviation used in The Unscrambler. / de Vries, S.; ter Braak, C.J.F.

    In: Chemometrics and Intelligent Laboratory Systems, Vol. 30, 1995, p. 239-245.

    Research output: Contribution to journalArticleAcademicpeer-review

    TY - JOUR

    T1 - Prediction error in partial least squares (PLS) regression: a critique on the deviation used in The Unscrambler

    AU - de Vries, S.

    AU - ter Braak, C.J.F.

    PY - 1995

    Y1 - 1995

    N2 - Partial least squares (PLS) regression is commonly used for multivariate calibration of instruments. Because of the need to know the quality of the prediction in a specific unknown sample and the lack of theory, an ‘empirically found formula’ to express the uncertainty is utilized in The Unscrambler II software, the de-facto standard in computer software for PLS. In this critique the formula is examined theoretically and by simulation. It is concluded that this formula underestimates the root mean squared error of prediction in most practical applications of PLS. A change of the formula is planned in the next version of The Unscrambler. In the mean time users of The Unscrambler ver 5.5 or lower should multiply the reported deviation by a factor of at least , to get a reasonable estimate of the prediction error

    AB - Partial least squares (PLS) regression is commonly used for multivariate calibration of instruments. Because of the need to know the quality of the prediction in a specific unknown sample and the lack of theory, an ‘empirically found formula’ to express the uncertainty is utilized in The Unscrambler II software, the de-facto standard in computer software for PLS. In this critique the formula is examined theoretically and by simulation. It is concluded that this formula underestimates the root mean squared error of prediction in most practical applications of PLS. A change of the formula is planned in the next version of The Unscrambler. In the mean time users of The Unscrambler ver 5.5 or lower should multiply the reported deviation by a factor of at least , to get a reasonable estimate of the prediction error

    KW - PLS

    KW - The Unscrambler

    KW - PLS

    U2 - 10.1016/0169-7439(95)00030-5

    DO - 10.1016/0169-7439(95)00030-5

    M3 - Article

    VL - 30

    SP - 239

    EP - 245

    JO - Chemometrics and Intelligent Laboratory Systems

    JF - Chemometrics and Intelligent Laboratory Systems

    SN - 0169-7439

    ER -