For the first position, funded by the Fund for Scientific Research (Flanders, Belgium), we are looking for candidates with a strong mathematical foundation who have conducted methodological research in statistics with the aim to advance data analysis, to work on the project entitled `Bias-reduced Double-robust estimation’ (Principal Investigator: Prof. Dr. Stijn Vansteelandt, project description given below). We particularly welcome promising candidates who have expertise in regularisation, model selection, semi-parametric inference and/or empirical processes. Pre-doctoral candidates with a strong mathematical foundation who are interested to work towards a PhD on this research project are also invited to express their interest.
For the second position, funded by the Ghent University Hospital, we are looking for candidates with a strong methodological interest and experience with collaborative research, to work on the project entitled `Causal analysis of large electronic health registers’ in collaboration with Prof. Dr. Stijn Vansteelandt from the Statistics research group in the Department of Applied Mathematics, Computer Science and Statistics, and Prof. Dr. Johan Decruyenaere, from the Intensive Care Outcomes Research Group at the Ghent University Hospital. We particularly welcome promising candidates who have expertise in causal inference, the analysis of electronic health records, or large-scale data analysis.
The research group on Statistics in the Department of Applied Mathematics, Computer Science and Statistics at Ghent University is internationally well known for its research in biostatistics, in particular, on causal inference, missing data, survival analysis and statistical genomics. The successful candidate for this position is expected to pursue innovating research at the highest international level, and to participate in limited teaching activities and student guidance. Both fellowships will start in 2016, but the starting date can be negotiated.
Applicants should have a Ph.D. in Statistics or a degree recognized as equivalent; they are expected to have a strong background in the mathematical foundations of statistics and show evidence of high-level research. To receive full consideration, applicants must submit their curriculum vitae, a brief (approximately 1/2 page) research statement (in English) and up to three selected publications or preprints electronically or by regular mail as soon as possible and preferably before March 1 to:
Department of Applied Mathematics, Computer Science and Statistics
Krijgslaan 281, S9
9000 Gent, Belgium
tel: ++32 9 2644776
Project description `Bias-reduced Double-robust estimation’: Most data analyses rely on statistical working models, which are not of primary scientific interest, but are needed to accommodate sparseness in the data. For instance, the estimation of treatment effects typically requires modelling the dependence of either outcome or treatment decisions on covariates. Misspecification of these models can be difficult to detect, yet seriously bias the treatment effect estimate. This is a major concern in practice, which has encouraged the growing popularity of double-robust data analyses. These rely on two working models, but demand only one of them to be correct, thereby giving the data analyst two chances of a valid analysis result. Bias- reduced double-robust (BR-DR) data analyses expand this double-protection property even further to realistic settings where both working models are misspecified. This is accomplished by fitting these working models in such a way that it minimises the (large sample) bias of the double-robust estimator of, e.g. the treatment effect. The theory on BR-DR estimation is novel and promising, especially considering the lack of statistical theories to control bias under model misspecification. In this project, we will develop greater insight into the properties of this method and of hybrid strategies that target bias-reduction in specific directions. We will extend it to make it widely applicable to all double-robust analyses, and we will develop its assimilation into regularisation and model selection methods.