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:Approximate Bayesian Computation for model selecti
on - Christian Robert\, Universite Paris-Dauphine
and IUF
DTSTART;TZID=Europe/London:20120127T160000
DTEND;TZID=Europe/London:20120127T170000
UID:TALK32950AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/32950
DESCRIPTION:Approximate Bayesian computation (ABC)\, also know
n as likelihood-free\nmethods\, has become a stand
ard tool for the analysis of complex models\,\npri
marily in population genetics but also for complex
financial models. The\ndevelopment of new ABC met
hodology is undergoing a rapid increase in the\npa
st years\, as shown by multiple publications\, con
ferences and even\nsoftwares. While one valid inte
rpretation of ABC based estimation is\nconnected w
ith nonparametrics\, the setting is quite differen
t for model\nchoice issues. We examined in Grelaud
et al. (2009) the use of ABC for\nBayesian model
choice in the specific of Gaussian random fields (
GRF)\,\nrelying on a sufficient property to show t
hat the approach was legitimate.\nDespite having p
reviously suggested the use of ABC for model choic
e in a\nwider range of models in the DIY ABC softw
are (Cornuet et al.\, 2008)\, we\npresent in Rober
t et al. (PNAS\, 2011) theoretical evidence that t
he general\nuse of ABC for model choice is fraught
with danger in the sense that no\namount of compu
tation\, however large\, can guarantee a proper ap
proximation\nof the posterior probabilities of the
models under comparison. In a more\nrecent work (
Marin et al.\, 2011)\, we expand on this warning t
o derive\nnecessary and sufficient conditions on t
he choice of summary statistics for\nABC model cho
ice to be asymptotically consistent.\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0W
B
CONTACT:Richard Samworth
END:VEVENT
END:VCALENDAR