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:MRC Biostatistics Unit Seminars
SUMMARY:"\;Some aspects in high-dimensional Bayesian m
odel choice"\; - Dr David Rossell\, University
of Warwick
DTSTART;TZID=Europe/London:20160426T143000
DTEND;TZID=Europe/London:20160426T153000
UID:TALK65253AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/65253
DESCRIPTION:Given a collection of candidate probability models
for an observed data y\, a fundamental statistica
l task is to evaluate which models are more likely
to have generated y. Tackling this problem within
a Bayesian framework requires one to complement t
he probability model for y (likelihood) with a pri
or probability model on the parameters describing
each of the candidate models\, as well as to speci
fy model prior probabilities and possibly a utilit
y function. We shall review some recent strategies
for high-dimensional model choice\, and then disc
uss what we denominate the "model separation princ
iple". This principle states that the models under
consideration should be minimally different from
each other\, else it becomes hard to distinguish t
hem on the basis of the observed y. In the common
setting where some of the models are nested this p
rinciple is violated\, as say Model 1 is a particu
lar case of Model 2 and thus these models are not
well separated. We shall review a class of prior d
istributions called non-local priors (NLPs) as a w
ay to enforce the model separation principle and s
ome of the NLP properties\, focusing on parsimony
and accelerated convergence rates in high-dimensio
nal inference. We shall illustrate their use in on
going work related to regression and robust regres
sion.
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Pub
lic Health\, University Forvie Site\, Robinson Way
\, Cambridge
CONTACT:Alison Quenault
END:VEVENT
END:VCALENDAR