=======================================================================
SEMINARS at the MRC BIOSTATISTICS Unit, Cambridge
EASTER TERM 2003
=======================================================================
29th April
Professor Simon Tavaré,
Department of Molecular and Computational Biology,
University of Southern California.
Title: Who needs likelihoods?
Abstract:
For problems involving complex stochastic models, the evaluation of
likelihoods is either extremely time consuming or impossible. This
raises the obvious question of how to do likelihood-based inference
when one does not have likelihoods. In this talk, I will illustrate
some simulation-based approaches for generating observations from
posterior distributions in such cases.
The approaches rely on the ability to simulate the underlying stochastic
model quickly (although not by estimating the likelihood). In addition to
examples using the rejection algorithm, I will
give one approach to Markov chain Monte Carlo that does not make use of a
likelihood ratio in the Hastings rejection step. Approximate sufficiency
plays a role in the method, and raises some interesting research problems.
The methods, which are accessible to any MSc level statistician,
will be illustrated by examples from inference in population genetics.
=======================================================================
20th May
Dr. Paddy Farrington,
Department of Statistics, The Open University.
Title: Statistical methods for the surveillance of vaccine-preventable
infections.
Abstract:
One aim of vaccination programmes is to maintain susceptibility
levels below the epidemic threshold. This means that an infectious
individual infects, on average, fewer than one other individual,
resulting in certain termination of the chain of transmission.
Interest thus focuses on the average number of individuals infected
by a single infectious person introduced into the population at
time t, R(t). R(t) is known as the effective reproduction number.
The aim of surveillance is to monitor R(t), and introduce additional
control measures if its value exceeds unity.
A common approach to estimating R(t) is based on deterministic
age-stratified transmission models, using data from surveys of
population immunity to estimate transmission parameters. I will
describe the method, and outline some new work to take account
of individual heterogeneity as well as age effects. If time allows
I will also discuss some issues relating to model uncertainty.
The talk will be illustrated with mumps and rubella data from
England and Wales.
=======================================================================
10th June
Dr. Elisabeth Allen,
Department of Statistical Science,
University College London.
Title: Multi-organ system rheumatological disease: statistical
analysis of outcome measures and their interrelationships.
Abstract:
Dermatomyositis, polymyositis and inclusion body myositis
(myositis) are usually regarded as a heterogeneous group of
autoimmune rheumatic diseases. To gauge more accurately the value of
conventional and newer therapies as they are introduced, a group of
specialists recently developed a set of measures for assessing
myositis outcomes. The design and analysis of two real patient exercises
carried out to study the proposed measures will be reported.
An approach to the study of reliability and agreement will be presented
and inference procedures for ratios of standard errors are developed.
These newly developed myositis measures are based on previous work in
systemic lupus erythematosus; a major autoimmune rheumatic disease.
International attempts to define validated disease activity and damage
indices to assess patients with lupus have provided a consistent way to
assess the disease.
There is however a need to better understand predictors of disease
activity to improve and standardise therapy and to prevent the
development of chronic damage. An analysis that attempts to examine
the interrelationships between disease activity in the different organ
systems will be presented. The analysis is based on logistic regression
methodology and the usefulness of separate logistic regressions with
dynamic covariates for the analysis of multinomial panel data is illustrated.
The efficiency of the approach relative to modelling disease activity in
continuous time is investigated.
=======================================================================
Seminars start at 2.30pm in the Large Seminar Room, 1st Floor,
Institute of Public Health, University Forvie Site, Robinson Way,Cambridge.
All welcome to attend.
=======================================================================
Daniela De Angelis
MRC Biostatistics Unit Tel: + (0)1223 330390
Institute of Public Health
Robinson Way Fax: + (0)1223 330388
Cambridge
CB2 2SR Web: www.mrc-bsu.cam.ac.uk
United Kingdom
|