zbMATH — the first resource for mathematics

On the asymptotic efficiency of directional models checks for regression. (English) Zbl 1383.62128
Ferger, Dietmar (ed.) et al., From statistics to mathematical finance. Festschrift in honour of Winfried Stute. Cham: Springer (ISBN 978-3-319-50985-3/hbk; 978-3-319-50986-0/ebook). 71-87 (2017).
Summary: In a landmark paper, W. Stute [Ann. Stat. 25, No. 2, 613–641 (1997; Zbl 0926.62035)] introduced smooth and directional tests for homoskedastic regression models based on weighted averages of estimated principal components of the CUSUM of residual concomitants process. This article shows that Stute’s [loc. cit.] directional test is asymptotically uniformly most powerful in a semiparametric context, considering the joint distribution of the regression error term and the covariate as a nuisance parameter. Moreover, the directional test is shown to be asymptotically equivalent to the classical \(t\)-ratio under homoskedasticity. This article also studies the heteroskedastic case, where the directional test based on the CUSUM of standarized residual concomitants [W. Stute et al., ibid. 26, No. 5, 1916–1934 (1998; Zbl 0930.62044)] is asymptotically equivalent to the semiparametric efficient test, which is the \(t\)-ratio using the generalized least squares estimator.
For the entire collection see [Zbl 1383.62010].

62G10 Nonparametric hypothesis testing
62G20 Asymptotic properties of nonparametric inference
62E20 Asymptotic distribution theory in statistics
Full Text: DOI