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:Hierarchical Evolutionary Stochastic Search with A
daptive Proposals - Leonardo Bottolo\, Institute o
f Mathematical Sciences\, Imperial College London
DTSTART;TZID=Europe/London:20080701T143000
DTEND;TZID=Europe/London:20080701T153000
UID:TALK11861AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/11861
DESCRIPTION:Multivariate regression models with many responses
has attracted the attention of the statistical co
mmunity in very recent years. A notable example is
the paradigm of eQTL analysis\, where thousands o
f transcripts are regressed versus (hundred of) th
ousands of markers. In this context the usual prob
lem of multimodality of the posterior distribution
\, when p>>n\, is further exacerbated by the dimen
sion of the response matrix\, usually q>>n. The pr
oblem can be even more complex when transcriptomic
data are available for multiple tissues\, introdu
cing a third dimension in the response matrix.\n\n
In this talk we introduce a new searching algorith
m called Hierarchical Evolutionary Stochastic Sear
ch (HESS) where the responses are linked in a hier
archical way. To reduce the computational burden\,
most of the regression parameters are integrated
out. A novel sampling strategy based on Evolutiona
ry Monte Carlo has been designed to efficiently sa
mple from the huge parametric space. Moreover the
whole set of past visited models are also consider
ed through an adaptive proposal distribution\, all
owing the algorithm to balance between the freedom
of exploration and the ability to persist on mode
ls in regions of high posterior probability.\n\nSi
mulated and real data sets are analysed to demonst
rate the performance of the proposed algorithm whe
n p and q are both larger than n and when multiple
tissues are considered. Collaborators on various
aspects of work: Sylvia Richardson\, David Welsh a
nd Enrico Petretto.
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Publ
ic Health\, University Forvie Site\, Robinson Way\
, Cambridge
CONTACT:Nikolaos Demiris
END:VEVENT
END:VCALENDAR