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:A flexible regression approach using GAMLSS - Mik
is Stasinopoulos\, London Metropolitan University
DTSTART;TZID=Europe/London:20100420T143000
DTEND;TZID=Europe/London:20100420T153000
UID:TALK23434AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/23434
DESCRIPTION:Generalized Additive Models for Location\, Scale a
nd Shape (GAMLSS) were introduced by Rigby and Sta
sinopoulos (2005). They refer to a very general re
gression type model in which both the systematic a
nd random parts of the model are highly \nflexible
and where the fitting algorithm is fast enough to
allow the rapid exploration of very large and com
plex data sets. GAMLSS is a general framework for
univariate regression type statistical problems. I
n GAMLSS the exponential family distribution assum
ption used in Generalized Linear Model (GLM) and G
eneralized Additive Model (GAM)\, (see Nelder and
Wedderburn\, 1972 and Hastie and Tibshirani\,\n199
0\, respectively) is relaxed and replaced by a ver
y general distribution family including highly ske
w and kurtotic discrete and continuous distributio
ns. The systematic part of the model is expanded t
o allow modelling not only the mean (or location)
but all the other parameters of the distribution o
f y as linear parametric\, non-linear parametric a
nd/or additive\n(smoothing) non-parametric functio
ns of explanatory variables and/or random effects
terms. Maximum (penalized) likelihood estimation i
s used to\nfit the models. For medium to large siz
e data\, GAMLSS allow \nflexibility\n in statistic
al modelling far beyond other currently available
methods. The GAMLSS framework is implemented in R.
\n\nThe most important application of GAMLSS up to
now is its use by the Department of Nutrition for
Health and Development of the World Health Organi
zation to construct the worldwide standard growth
centile curves\,\nsee WHO(2006). The range of poss
ible applications for GAMLSS is very wide and exam
ples will be given of its usefulness in modelling
data.\nIn the talk we will describe the GAMLSS mod
el\, the variety of different (two\, three and fou
r parameter) distributions that are implemented wi
thin\nthe R GAMLSS package and the variety of diff
erent additive (smoothing) terms that can be used.
New distributions and new additive terms can be\n
added easily to the R package. The use of censored
data\, truncate distributions and finite mixture
of distributions within the GAMLSS framework\, wil
l also be described.\n\n*References*\n\n\nHastie\,
T.J.\, and Tibshirani\, R.J. (1990) Generalized A
dditive Models. London: Chapman & Hall.\n\n\nIhaka
\, R.\, and Gentleman\, R. (1996)\, A Language for
Data Analysis and Graphics\, Journal of Computati
onal and Graphical Statistics\, 5\,3\,299-314.\n\n
\nNelder\, J.A. and Wedderburn\, R.W.M. (1972) Gen
eralized Linear Models. J. R. Statist. Soc. A\, 13
5\, 370-384.\n\n\nRigby\, R.A. and Stasinopoulos\,
D.M. (2005) Generalized Additive Models for Locat
ion\, Scale and Shape (with discussion). Appl. Sta
tist.\, 54\, 1-38.\n\n\nWHO Multicentre Growth Ref
erence Study Group (2006) WHO Child Growth Standar
ds: Length/height-for-age\, weight-for-age\, weigh
t-for-length\, weight-for-height and body mass ind
ex-for-age: Methods and development. Geneva: World
Health Organization.
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Publ
ic Health\, University Forvie Site\, Robinson Way\
, Cambridge
CONTACT:Michael Sweeting
END:VEVENT
END:VCALENDAR