The RSS Statistical Computing Section announce a half day meeting on
nonlinear mixed-effects modelling.
Date & Time: Wednesday 12th May 2004, 2:00-5:00 pm
Place: Errol Street
Refreshments: Tea will be served at 3:45pm.
Enquiries: Peter Lane +44 (0)1279 646558
"The use of nonlinear mixed-effects modelling in pharmacokinetics and
pharmacodynamics"
Leon Aarons (School of Pharmacy and Pharmaceutical Sciences, University of
Manchester)
Synopsis: In this presentation I will introduce population harmacokinetics
and pharmacodynamics (pk/pd), particularly in the context of drug
evelopment. The two features of population pk/pd that makes data analysis
particularly challenging are: 1) the models that describe this type of
data are nonlinear in the parameters; and 2) the data is generally very
sparse. Consequently nonlinear mixed effects modelling has been
extensively used in this area, generally by statistically lay users.
However nlme analysis is by no means trivial. The talk will be illustrated
with a number of examples that illustrate different aspects of nlme
modelling.
"Non-normal nonlinear mixed-effects modelling using the SAS procedure
NLMIXED"
Mike Patefield (School of Applied Statistics, University of Reading)
Synopsis: PROC NLMIXED fits nonlinear mixed models by maximum likelihood
using numerical integration over the normally distributed random effects
to form the likelihood of the response. There is complete flexibility over
the choice of distribution of the response. The approach is illustrated by
examples fitting generalised linear and nonlinear mixed models to repeated
binary and categorical data. For generalised linear models comparisons are
made with the MLwiN package. Some of the pitfalls arising using NLMIXED
are discussed together with methods of overcoming them.
"Software for nonlinear mixed-effects modelling"
Mike K Smith (Pfizer Global R&D, Sandwich )
(NOTE: this talk replaces one by Dave Lunn who unfortunately has had to
withdraw)
Synopsis: I will give some background information on what nonlinear mixed-
effects models are, why they can be useful in analysing repeated measures
and what makes analysing such data challenging. Various software packages
and routines are available to perform this kind of analysis, and I will
compare and contrast these.
|