Dr Hugh Miller
2011 Chancellor’s Award Winner (Science and Engineering)
The Chancellor’s Prize for Excellence in the PhD Thesis is awarded on an annual basis (in each of the Faculties grouping of: Humanities and Creative Arts, Social Sciences, Science and Engineering and Medicine, Dentistry and Health Sciences).
Dr Hugh Miller is the Chancellor’s Prize Award Winner in Science and Engineering for his thesis entitled “Statistical Methods for the analysis of high-dimensional data”, completed under the supervision of Prof Peter Hall.
High-dimensional statistics is a fast-growing research area, due to a large number of modern problems characterised by an excess of potentially useful variables. New statistical research is required to understand whether traditional analyses are appropriate, or what approaches should be adopted if they are not.
Dr Miller addressed a number of topics to improve statistical methodology to help understand these types of problems. In particular he investigated:
- How to identify important features and test variable relationships, particularly when important nonlinearities may be present in the data;
- How to create nonparametric predictive models that adapt to the observed variable importance;
- How to assess the reliability of a ranking of available variables, and to determine circumstances where parts of these rankings are expected to be reliable;
- How to perform hypothesis tests on the extremes of a population; and
- How to build a multivariate predictive model directly on high-dimensional data while simultaneously protecting against over-fitting