@krzysztofsztompka/garch
    Preparing search index...

    Type Alias DieboldMariano

    Is one forecast's loss lower than another's by more than noise? The Diebold–Mariano test on the per-origin loss differences, with the Harvey–Leybourne–Newbold small-sample correction.

    type DieboldMariano = {
        df: number;
        lags: number;
        meanDifference: number;
        observations: number;
        pValue: number;
        standardError: number;
        statistic: number;
    }
    Index
    df lags meanDifference observations pValue standardError statistic
    df: number
    lags: number

    Autocovariance lags in the HAC variance: horizon − 1.

    meanDifference: number

    mean(lossA − lossB): below 0, A lost less.

    observations: number
    pValue: number

    Two-sided: is either forecast better?

    standardError: number

    The HAC standard error of that mean.

    statistic: number

    The HLN-corrected statistic, referred to Student's t with df degrees of freedom.

    Settings
    Member Visibility

    On This Page
    Properties