BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Statistics
SUMMARY: Stepwise Searching for Feature Variables in High-
 Dimensional Linear Regression -  Qiwei Yao\, LSE
DTSTART;TZID=Europe/London:20090130T160000
DTEND;TZID=Europe/London:20090130T170000
UID:TALK16634AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/16634
DESCRIPTION:We investigate the classical stepwise forward and 
 backward\nsearch methods\nfor selecting sparse mod
 els in the context of linear regression with the\n
 number of candidate variables p greater than the n
 umber of observations\nn. Two types of new informa
 tion criteria BICP and BICC are proposed to\nserve
  as the stopping rules in the stepwise searches\, 
 since the\ntraditional information criteria such a
 s BIC and AIC are designed for the\ncases with p<n
 \, and may fail spectacularly when p is close to o
 r\ngreater than n.  The proposed methods are illus
 trated in a simulation\nstudy which indicates that
  the new methods outperform a counterpart LASSO\ns
 elector.  The consistency of the stepwise search i
 s investigated when\nboth $n$ and p tend to infini
 ty. We show that a stepwise forward\naddition foll
 owed by a stepwise backward deletion\, both contro
 lled by a\nversion of BICP\, leads to a consistent
  estimated model under the sparse\nRiesz condition
 .\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0W
 B
CONTACT:
END:VEVENT
END:VCALENDAR