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:Statistical Laboratory Graduate Seminars
SUMMARY:D-optimal design of experiments - Radoslav Harman\
, Comenius University
DTSTART;TZID=Europe/London:20111124T150000
DTEND;TZID=Europe/London:20111124T160000
UID:TALK33839AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/33839
DESCRIPTION:Consider an experiment consisting of a series of l
inear regression trials. The question is: How shou
ld we choose the trials in order to obtain "as muc
h information as possible" from the experiment? Th
e most classic answer is to choose the trials such
that the resulting confidence ellipsoid for the u
nknown regression parameter has minimum achievable
volume\, which leads to the so-called D-optimal d
esign. In the talk we will define the basic notion
s of exact and approximate designs and prove the f
undamental "equivalence theorem" for D-optimality.
We will also give several concrete examples of D-
optimal exact and approximate designs for various
linear regression models\, relating the design pro
blems to areas as different as the theory of Brown
ian motion and the number of spanning trees of gra
phs.
LOCATION:CMS\, MR9
CONTACT:Elena Yudovina
END:VEVENT
END:VCALENDAR