VIEWS: 13 PAGES: 1 POSTED ON: 2/27/2012
MRC Capacity Building Research Studentship Biostatistics Unit, Centre for Epidemiology and Biostatistics, faculty of Medicine and Health, University of Leeds Improved statistical modelling for observational research: understanding and addressing the problems associated with collinearity Supervisors: Professor Mark S Gilthorpe, Dr Paul D Baxter, Dr Yu-Kang Tu Sponsor: MRC / University of Leeds Email: firstname.lastname@example.org Tel: 0113 343 7497 The Biostatistics Unit, Centre for Epidemiology and Biostatistics, has the following prestigious (3+1 years) full-time studentship available from 1st September 2008, funded by the Medical Research Council. This capacity-building studentship will support a 1-year MSc in Health Research, followed by a 3-year PhD in Statistical Epidemiology. Collinearity amongst covariates in linear regression models has long been recognised as a potential source of bias. Various ‘solutions’ have been proposed, though one important issue almost entirely omitted in the current literature is the importance of the relationship between the outcome and correlated covariates. Using vector geometry, it can be shown that the impact of collinearity on the model, such as changes in regression coefficients, cannot be judged by the correlation structure of the covariates alone – their relationship with the outcome is crucial. Traditional diagnostics of collinearity are thus insufficient in evaluating adverse effects or model instability. The challenges for the student are to develop new diagnostic tools to evaluate the extent of collinearity in statistical models, both beneficial (appropriate adjustment for confounding) and adverse (bias). For generalized linear models, further theoretical work is needed to develop a statistical index to characterize associations between the outcome and all collinear covariates. Theoretical derivations and computer simulations are needed to evaluate the extent of collinearity. The student is then required to implement these new methods in statistical packages, such as R or Stata, to provide a clear illustration to general users of statistics how to quantify beneficial and adverse collinearity in their statistical model. Simulations and genuine data will be used to test the validity of these new methods. Most covariates in epidemiology and public health are collinear, and the implementation of these methods will enable biomedical researchers to develop more appropriate models to provide robust research findings. This studentship is therefore anticipated to have a huge impact on the statistical modelling undertaken in these fields. Students are required to have at least an upper second class degree in mathematics, statistics or related subject, and an interest in applied methodology in biomedical research. The candidate will receive basic training towards an MSc in Health Research, involving basic epidemiology and statistical epidemiology. Informal enquires regarding this studentship can be directed to either Professor Gilthorpe (email@example.com; Tel 0113 343 1913) or Dr Baxter (firstname.lastname@example.org; Tel 0113 343 5162). Further details about the research interests of the group can be found at http://www.leeds.ac.uk/medhealth/light/research/deb/index.html. Candidates must be eligible for UK/EU fee status (EU students MUST have been in the UK for at least three years prior to commencement of studentship, or being ‘migrant workers’ at the time of application). The studentship will cover full fees for UK/EU students, a research training support grant, conference allowance, and a maintenance stipend (approximately £14,600 per year). Applicants should send a statement of interest (no more than 1 page) and a CV with the names and contact details of two academic referees to Miss Claire Walton (Postgraduate Research Student Co-ordinator, Faculty Graduate School, Room 10.110, Level 10, Worsley Building, Leeds LS2 9NL) by Wednesday 30th April 2008.
Pages to are hidden for
"Collinearity has long been recognised as a potential problem in "Please download to view full document