Testing in robust ANOVA

Henry I. Braun, Don McNeil

    Research output: Contribution to journalArticlepeer-review

    1 Citation (Scopus)


    The bisquare analysis of a simple linear model using an iteratively reweighted least squares (IRLS) algorithm is considered. A standard weighted least squares analysis of variance using the final configuration of weights is employed to derive a robust analysis of variance. The small sample properties of the resulting pseudo F-statistic are explored by means of experimental sampling. Employing six error distributions and nine configurations of effects, the robustness of validity and power characteristics of the statistic are established. In addition, the problem of nuisance effects and the role of the noncentrality parameter are investigated. Approximating the distribution of the statistic by an F-distribution proves quite successful, with the chief difficulty being the correct assignment of the degrees of freedom.
    Original languageEnglish
    Pages (from-to)149-165
    Number of pages17
    JournalCommunications in Statistics - Theory and Methods
    Issue number2
    Publication statusPublished - 1981


    • Robust F-test
    • robustness of validity
    • power
    • non-centrality parameters


    Dive into the research topics of 'Testing in robust ANOVA'. Together they form a unique fingerprint.

    Cite this