PhD STUDENTSHIPS, MRC BIOSTATISTICS UNIT, CAMBRIDGE
The BSU is an internationally recognised research unit specialising in
statistical modelling with application to medical, biological or public health
sciences. Details of the work carried out in the Unit appear on the website
www.mrc-bsu.cam.ac.uk. The Unit has 3 MRC PhD studentships to start from October
2005. Awards cover Cambridge University fees and a stipend of at least 10,500
pounds for a period of 3 years. Applicants must have or expect to get a first or
high 2.1 honours degree in mathematics, statistics or a related discipline. A
masters degree is desirable but not essential.
Examples of some of the projects available for PhD study are given below. Other
student-proposed projects will be considered.
Full research council funding can be awarded to UK citizens, fees only are
payable for citizens of EU countries. It is not possible to award studentships
to citizens of countries outside Europe. Non-UK citizens with their own source
of funding are encouraged to apply.
Interested applicants should send CV, letter and contact details of 2 academic
referees to Linda Sharples at the Biostatistics Unit by 28th February
2005.E-mail applications are acceptable to [log in to unmask]
PROJECTS
Title: Modelling within-individual variation in epidemiological data
Supervisor: Ian White
Many risk factors for diseases are subject to measurement error and
within-individual variation. This has the effects (i) that associations with
disease risk are underestimated ("regression dilution") and (ii) that adjustment
for confounders is inadequate ("residual confounding").
The magnitude of within-individual variation may be estimated if a second risk
factor measurement is available at least for some individuals, and corrections
for regression dilution and residual confounding may be applied. Commonly used
methods make a number of assumptions about the relationship between the "true"
risk factor and the outcome, about the regression of the second measurement on
the first, and about the impact of other variables.
This project will make use of a dataset such as a meta-analysis of major cohort
studies of cardiovascular disease with individual participant data to explore
the following issues:
· the validity of the assumptions of commonly used correction methods
· the impact of violations of these assumptions
· developing new methodology for such data
· allowing for heterogeneity between the studies of the meta-analysis
The work will be supervised by Ian White, and advisors will include Simon
Thompson, Angela Wood and a senior epidemiologist.
Title: Using and reporting logistic regression
Supervisor: Ian White
Logistic regression is a powerful and commonly used tool in the analysis of
clinical and epidemiological data. This project will explore one or more of the
following three issues in the use of logistic regression in observational
studies and its presentation to non-statisticians.
Firstly, many methods are known for quantifying the predictive ability of a
logistic regression model, but none of them is immediately meaningful to
clinical audiences. The choice of method is likely to differ between diagnostic
problems where a good model is one that correctly identifies cases, and
prognostic problems where a good model is merely one that identifies a high-risk
group. This project will compare various existing methods [1], paying attention
to the differences between with-sample and out-of-sample prediction, and will
attempt to devise new clinically relevant methods. For example, if treatment
were available for high-risk individuals then it would be possible to quantify
the mean health benefit of a given predictive model.
Secondly, presentation of logistic regression results in terms of likelihood
ratios is helpful to clinical audiences. This part of the project will compare a
newly proposed method [2] with previously available methods and will further
develop the method, for example to deal with zero counts.
Thirdly, quantitative variables are often entered into logistic and other
regression models in a grouped from, for example in tertile groups. This is
known to entail a loss of power. This part of the project will both quantify the
loss of power and consider the possible benefits of grouping. When a number of
studies explore associations between disease and a quantitative outcome, some
may use regression on a continuous variable while others may use different
numbers of groups with different cut points. The project will explore methods
for the meta-analysis of such studies from reported data.
Part of the project will utilise a large obstetric database in collaboration
with Gordon Smith, Professor of Obstetrics and Gynaecology. Data from other
epidemiological studies will also be available. Simon Thompson will be a
statistical advisor.
1. M. Mittlböck and M. Schemper. Explained variation in logistic
regression. Statistics in Medicine 1996;15:1987-1997.
2. Smith GCS, Dellens M, White IR, Pell JP. Combined logistic and Bayesian
modeling of cesarean section risk. Am J Ob Gyn 2004;191:2029-34.
Title: Modelling longitudinal performance indicators and retrospectively
identifying 'odd' performance
Supervisor: Dr. David Spiegelhalter
Performance indicators such as mortality rates are collected over time for
hospitals and other institutions and, for example, the Healthcare Commission
uses them as a guide for assessing and inspecting the NHS in England. It is
important to be able to model longitudinal performance data and characterise
outlying periods or institutions, as well as
assessing change. Hierarchical models, either likelihood-based or fully
Bayesian, have been suggested as a basic methodology, but their properties and
the precise methods to be used for characterising 'oddness' are still an open
question. Somewhat complex schemes
suggested in the US require further investigating. A range of datasets are
available for analysis, including live-birth rates in IVF clinics and MRSA rates
in hospitals.
Title: Longitudinal studies of chronic diseases with applications to
rheumatology
Supervisor: Prof. Vern Farewell
This studentship will deal with methods for the analysis of longitudinal
rheumatological databases. One focus will be on comparative methods for
prognosis studies with an emphasis on prediction of the future course of
disease. Both clinical and quality of life measures of current disease state
are of interest. The use of database information for cost-effectiveness
studies may also be explored.
Title: Sequential methods for meta-analysis
Supervisor: Dr. Julian Higgins
Meta-analyses are statistical analyses that integrate the findings of multiple
studies. They are commonly encountered as part of retrospective, systematic
reviews of existing research. However, they are increasingly being considered in
prospective contexts. For example, if a particular epidemiological association
is identified as potentially important, data collections from multiple
epidemiological studies can be interrogated as part of a prospective
meta-analysis. This approach is now common in the field of human genome
epidemiology in particular, where interesting findings are confirmed or refuted
by attempting to replicate them systematically in other data sets. Similar
theoretical issues arise when meta-analyses are regularly updated, such as those
in the Cochrane Database of Systematic Reviews.
If meta-analyses are repeated using conventional statistical methods then the
risk of false-positive findings increases. Sequential methods for meta-analysis
are required that mirror established sequential methods for clinical trials. The
ability to incorporate random effects across studies, and to examine study-level
effect modifiers, are important aspects of meta-analysis methodology. There is
little, or no, suitable methodology for tackling these in the context of a
sequential meta-analysis.
This project will examine Bayesian and classical methods for sequential
meta-analysis, with applications in clinical trials and epidemiology,
particularly human genome epidemiology. The work will build on collaborative
work previously undertaken by Dr Higgins and Prof Anne Whitehead (The University
of Reading) and with Prof Deborah Ashby (Queen Mary, University of London). It
will involve collaboration with the Public Health Genetics Unit and the MRC
Epidemiology Unit in Cambridge, who are planning a series of sequential
meta-analyses of gene-disease association studies in type 2 diabetes, as well as
continuing collaboration with colleagues in Reading and London.
Title:Analysis of genome wide association studies
Supervisor: Dr. Frank Dudbridge
With the completion of the human genome sequencing project, many large scale
projects are now underway to unravel the genetic basis of common diseases. To
analyse these studies effectively, new methods are required to exploit the
complexity of the underlying genetics while managing the genomewide scale of the
data. This project will explore
methods for finding associations from multiple hypothesis tests, for integrating
Bioinformatics knowledge with genetic epidemiology, and for efficient analysis
of multiple-cohort studies. The work will be motivated by data from the
European Bloodomics consortium, which is identifying genetic risk factors in
coronary artery disease.
Title: Issues in combining data sources in cost-effectiveness analysis
Supervisor: Dr. Linda Sharples
Health service policy decisions are based on synthesing evidence from clinical
trials, meta-analyses of clinical trials, observational studies, registries and
population characteristics. A unified approach to modelling cost-effectivess
parameters is required including impact of varying assumptions about model
inputs and structure. Bayesian methodology is well suited to this type of
analysis. Datasets from a number of examples in cardiothoracic surgery and
medicine are available as examples. Dr. Nikolaos Demiris will act as an advisor
for this project.
Title: Mixing matrix models for the estimation of HIV incidence, prevalence,
and diagnosis rate, based on Bayesian synthesis of routine data.
Supervisor: Daniela De Angelis
The objective of this PhD is to develop evidence-based statistical models
of the prevalence, incidence and diagnosis rate of HIV infection in the UK.
Bayesian multi-parameter evidence synthesis will used to estimate all these
parameters simultaneously, using aggregated data from all the available
routine surveillance sources and from ad hoc surveys. The consistency of
the evidence sources with each other will also be assessed. Methods will
be developed for estimating both the incidence of HIV in specific risk
groups, and the rate at which individuals enter and leave high risk
groups. A further development will be estimation of a 'mixing matrix'
to express incidence in terms of risk-group specific prevalences.
The project will require an understanding of epidemiological data,
HIV surveillance sources, as well as mathematics and programming skills.
This is a collaborative project between the MRC Biostatistics Unit Cambridge
and the HIV & STD Division, Health Protection Agency, Colindale.
The PhD will be jointly supervised by Daniela De Angelis, Tony Ades,
Graham Medley, and Noel Gill.
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
|