BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayes and empirical-Bayes multiplicity adjustment in the variable-
 selection problem - James Scott\, Duke University
DTSTART:20090313T160000Z
DTEND:20090313T170000Z
UID:TALK16640@talks.cam.ac.uk
CONTACT:8419
DESCRIPTION:In this talk\, I will present a theorem that characterizes a\n
 surprising\ndiscrepancy between fully Bayes and empirical-Bayes\napproache
 s to\nmultiplicity adjustment in linear regression.  This\ndiscrepancy ari
 ses\nfrom a different source than the failure to account for\nuncertainty 
 in\nthe empirical-Bayes estimate\, which is the usual issue in\nsuch\nprob
 lems.  Indeed\, I will show that even at the extreme\, when\nthe\nempirica
 l-Bayes estimate converges asymptotically to the true\nparameter value\, t
 he potential for a serious difference\nremains.\n\nI will also highlight s
 ome interesting examples of Bayesian\nmultiplicity adjustment on large dat
 a sets\, with particular\nattention\nto a business application that involv
 es large-scale screening\nof\nfunctional data.\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR