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:Marginal Models for Dependent\, Clustered\, and Lo
ngitudinal Categorical Data - Wicher Bergsma\, Lon
don School of Economics and Political Science
DTSTART;TZID=Europe/London:20120203T160000
DTEND;TZID=Europe/London:20120203T170000
UID:TALK35235AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/35235
DESCRIPTION:In the social\, behavioural\, educational\, econom
ic\, and biomedical sciences\,\ndata are often col
lected in ways that introduce dependencies in the\
nobservations to be compared. For example\, the sa
me respondents are\ninterviewed at several occasio
ns\, several members of networks or groups are\nin
terviewed within the same survey\, or\, within fam
ilies\, both children and\nparents are investigate
d. Statistical methods that take the dependencies
in\nthe data into account must then be used\, e.g.
\, when observations at time one\nand time two are
compared in longitudinal studies. At present\, re
searchers\nalmost automatically turn to multi-leve
l models or to GEE estimation to deal\nwith these
dependencies. Despite the enormous potential and a
pplicability of\nthese recent developments\, they
require restrictive assumptions on the\nnature of
the dependencies in the data. The marginal models
of this talk\nprovide another way of dealing with
these dependencies\, without the need for\nsuch as
sumptions\, and can be used to answer research que
stions directly at\nthe intended marginal level. T
he maximum likelihood method\, with its\nattractiv
e statistical properties\, is used for fitting the
models.\n\nIn the talk I will also spend time on
some interesting mathematical problems\nthat arise
in marginal modelling. These relate to smoothness
of the manifold\nof probability distributions sat
isfying a marginal model\, and the linear\nindepen
dence of a set of marginal parameters. This is imp
ortant for the\napplicability of standard asymptot
ic theory\, and for the calculation of the\ncorrec
t number of degrees of freedom for a model.\n\nThi
s talk is based on a recent book by the authors in
the Springer series\nStatistics for the Social Sc
iences\, see www.cmm.st\n\nJoint work with Marcel
Croon and Jacques Hagenaars
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0W
B
CONTACT:Richard Samworth
END:VEVENT
END:VCALENDAR