Title: Developing, evaluating and applying methods for robustly handling dropout in cohort studies
Overview: The University of Bristol is offering a 3.5-year full time PhD in research around Population Health to start in 2020. This studentship is funded through the GW4BioMed MRC Doctoral Training Partnership. It consists of full UK/EU tuition fees, as well as a Doctoral Stipend matching UK Research Council National Minimum (£15,009 p.a. for 2019/20, updated each year).
Additional research and training funding are available over the course of the programme. This will cover costs such as research consumables, courses, conferences and travel. Additional competitive funds are available for high-cost training/research.
The studentship is based at the MRC Integrative Epidemiology Unit (MRC IEU) hosted by the Bristol Medical School. The MRC IEU provides a stimulating and supportive environment for PhD students in which to undertake world-class research. The student will also work with Dr Bartlett in the Department of Mathematical Sciences, University of Bath. For further information about the MRC IEU please see the website below.
http://www.bristol.ac.uk/integrative-epidemiology/
Project description: Cohort studies, (e.g., UK Biobank) are used to inform about clinical practice but face problems of missing data arising through non-response and dropout. The subsequent loss of information can undermine the validity of research results (i.e., biased analyses and loss of power). Data are classified as missing not at random (MNAR) when the observed data cannot explain all systematic differences between the observed and missing data. This MNAR property cannot be tested based on the observed data only but is a plausible scenario for many studies. Commonly used statistical methods are usually not applicable when data are MNAR.
Developing methods that allow for MNAR is a growing research area, especially methods that non-technical analysts (such as medical researchers and epidemiologists) can easily use. This PhD will use the latest research in Bayesian sensitivity analysis methods and instrumental variables for missingness to develop two distinct methods for analysing MNAR data. These methods will be developed in the context of missingness and dropout from UK Biobank and ALSPAC (the Avon Longitudinal Study of Parents and Children), taking advantage of linkage to routine data (from GP records and education) to provide some information about the missing data.
Bayesian sensitivity analyses aim to use external information and/or expert opinion to correct for MNAR data. Typically, such MNAR sensitivity analyses have involved analysts specifying priors for sensitivity parameters which are difficult to interpret and hence also difficult to elicit prior beliefs or knowledge about. In the second approach, the instrumental variable method uses variables (known as instruments) that are likely to be associated with missingness, but not with the incomplete variables themselves, to estimate the distribution of the complete data. Obtaining valid instruments can be challenging.
The main aims of this PhD will be: 1) Extend the latest Bayesian models to use routine linked data and develop simpler Bayesian approaches which use estimates of population quantities. 2) Investigate the robustness of instrumental variable methods for missingness and the potential to use instruments derived from linked data. 3) Conduct detailed comparisons of these developed methods with the best available methods. This PhD will make a major contribution by developing methods that will be of great relevance to all studies with incomplete data, especially much used resources such as UK Biobank.
Prospective students will be highly numerate; the project will include training in missing data concepts and general missing data methods, Bayesian analysis, instrumental variable analysis, and simulation studies. The PhD will result in high impact journal articles, and results being presented at national/international meetings and conferences.
Candidate requirements: Applications are welcome from high performing individuals across a wide range of disciplines closely related to natural sciences, biostatistics, genetics, biochemistry, mathematics and computer science who have, or are expected to obtain, a 2.1 or higher degree. Applications are particularly welcome from individuals with a relevant research master’s degree.
How to apply: You can apply for the studentship through the MRC GW4 BioMed DTP web site (https://www.gw4biomed.ac.uk/doctoral-students/) until 5pm on Monday 25th November 2019.
For informal enquiries contact Rachael Hughes: [log in to unmask]
Rachael Hughes
Senior Research Fellow
Population Health Sciences
Bristol Medical School
University of Bristol
Canynge Hall
39 Whatley Road
Bristol, BS8 2PS
+44 (0) 117 9287244
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|