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:Isaac Newton Institute Seminar Series
SUMMARY:Designs for mixed models with binary response - Wa
ite\, T (Southampton)
DTSTART;TZID=Europe/London:20110809T160000
DTEND;TZID=Europe/London:20110809T164500
UID:TALK32306AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/32306
DESCRIPTION:For an experiment measuring a binary response\, a
generalized linear model such as the logistic or p
robit is typically used to model the data. However
these models assume that the responses are indepe
ndent. In blocked experiments\, where responses in
the same block are potentially correlated\, it ma
y be appropriate to include random effects in the
predictor\, thus producing a generalized linear mi
xed model (GLMM). Obtaining designs for such model
s is complicated by the fact that the information
matrix\, on which most optimality criteria are bas
ed\, is computationally expensive to evaluate (ind
eed if one computes naively\, the search for an op
timal design is likely to take several months). \n
When analyzing GLMMs\, it is common to use analyti
cal approximations such as marginal quasi-likeliho
od (MQL) and penalized quasi-likelihood (PQL) in p
lace of full maximum likelihood. In this talk\, we
consider the use of such computationally cheap ap
proximations as surrogates for the true informatio
n matrix when producing designs. This reduces the
computational burden substantially\, and enables u
s to obtain designs within a much shorter time fra
me. However\, other issues also need to be conside
red such as the accuracy of the approximations and
the dependence of the optimal design on the unkno
wn values of the parameters. In particular\, we ev
aluate the effectiveness of designs found using th
ese analytic approximations through comparison to
designs that are found using a more computationall
y expensive numerical approximation to the likelih
ood.\n
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:Mustapha Amrani
END:VEVENT
END:VCALENDAR