Autore:
De Martini, Daniele Titolo:
Smoothed bootstrap for the correlation model with a locally orientated kernel using MCMC.Periodico:
Università degli Studi del Piemonte Orientale 'A. Avogadro' : Facoltà di Economia - Dipartimento di Scienze Economiche e Metodi Quantitativi "SEMEQ" - QuaderniAnno:
2002 - Volume:
02 - Fascicolo:
31 - Pagina iniziale:
1 - Pagina finale:
24This paper deals with the smoothed version of the bootstrap approach to the correlation model y = Xâ + å, where (y,X) ~ F and F is a distribution over Rk. This setting could improve classical bootstrap performances and its consistency through the convergence in Mallows metric of the kernel multivariate estimator of F is shown. Moreover, in order to obtain a better fitting of F, a locally orientated kernel smooth estimator, which also converges to F, is defined. The percentile-bootstrap and the hybrid-bootstrap can be applied to compute confidence intervals for â. A simulation study and a clinical application are presented, comparing classical bootstrap and dif-ferent smoothed bootstrap techniques. MCMC methods are used to resample from kernel multivariate distributions. The smoothed bootstrap looks preferable as it re-spects continuity hypothesis and avoids some computational problems, but its results are too conservative. The smoothed bootstrap with a locally orientated kernel main-tains these properties and provides more reliable results.
Esportazione dati in Refworks (solo per utenti abilitati)
Record salvabile in Zotero