Centre for Paediatric Epidemiology & Biostatistics
Institute of Child Health, University College London
Departmental seminar on Wednesday 10th March at 2pm
Room B Wellcome Trust Building, Institute of Child Health
30 Guilford Street, London WC1N 1EH
(Nearest tube station: Russell Sq)
Directions:
http://www.ich.ucl.ac.uk/gosh_families/your_visit/getting_here/index.html
Smooth centile curves for skew and kurtotic data
Bob Rigby & Mikis Stasinopoulos,
Statistics, OR and Mathematics (STORM) Research Centre,
London Metropolitan University
Abstract:
A general approach to smooth centile estimation is presented for a
continuous response variable
(e.g. head circumference) as a function of an explanatory variable (e.g. age),
where the response variable exhibits both skewness and kurtosis.
The model assumes independent observations (e.g. a cross sectional study).
The distribution of
the response variable is selected from several alternative four parameter
distributions
(e.g. Rigby and Stasinopoulos 2004a, 2004b). The four parameters of the
distribution relate to
location, scale, skewness and kurtosis respectively.
Smooth centile curves are obtained by modelling each of the four parameters
of the distribution
as either parametric (e.g. fractional polynomial) or nonparametric (e.g.
cubic smoothing spline)
functions of the explanatory variable.
Model selection and diagnostics are discussed. An example is presented from
the Fourth Dutch Growth
Study where centiles for head circumference against age are obtained.
The model used for smooth centile estimation is a special case of the
Generalized Additive Model
for Location, Scale and Shape (GAMLSS) of Rigby and Stasinopoulos (2004c).
The software for fitting GAMLSS, implemented as a package in the
statistical environment R
(all free of charge), will also be described.
The methodology and software are currently being applied by the World
Health Organisation to
centile estimation in their Multicentre Growth Reference Study (MGRS).
_______________________
Mario Cortina Borja
Senior Lecturer in Statistics
[log in to unmask] Phone +44(0)20 7905 2113 Fax +44(0)20 7905 2381
Paed. Epid. & Biostats, Institute of Child Health, UCL, London WC1N 1EH, UK
|