PhD applicants for 2020 entry please note: I have no capacity for new PhD students in 2020.
Please do not name me as potential supervisor in your application for PhD entry in 2020-21.
With apologies -- David Firth
Broad areas of PhD supervision
I have research interests in many areas of statistical methodology, especially likelihood theory and tools for statistical modelling. I also work on substantive applications that demand some methodological innovation: applied areas I have worked in previously are politics, sociology, public health, animal behaviour, bibliometrics, sport, etc. I am open to interesting statistical problems from any research discipline or other area of work.
Potential PhD students wishing to work inder my supervision are very welcome to suggest their own projects or research areas, bearing in mind my research interests as described above.
The list of topics below gives some specific ideas for PhD-project areas in which I would be happy to supervise a student with appropriate statistical background and interests, during the next few years.
David Firth. Last updated in September 2018.
Some specific PhD project areas of immediate interest
- Compositional data analysis. This is a well established area of multivariate statistical methodology, with applications in very many applied fields (including chemistry, geology, geography, politics, social science, economics, among others). The traditional data-analytic approach dates back to seminal work of Aitchison and co-authors in the 1980s, and is based on log-ratio transformation of the relative proportions that make up each observed composition. A PhD student working in this area would contribute to the development of a more robust approach that avoids explicit data-transformation. Some specific topics suitable for PhD study are models for compositional time series, models for classification and clustering of compositions, etc. A suitable student would have a strong grounding in statistical theory and methods, and an interest in computation for statistical models.
- Statistical learning from sequential pair-comparison data. The use of pair-comparison data, or more generally group-comparison data, to rank objects or to construct models that explain preferences, is common in many application fields, ranging from psychology through economics to market research and large-scale recommender systems, and also scientific fields as diverse as molecular biology and animal behaviour. The most commonly used statistical models are variants of the Bradley-Terry or Thurstone models, both of which are generalized linear models. This project will consider efficient, robust and targetted strategies for data acquisition and analysis when pairs or groups of objects are observed in sequence --- as often happens in practice. A suitable student would have a strong grounding in statistical theory and methods, and an interest in computation for statistical models.
- Modelling citation life-histories, and predicting trajectories of published research. Every scientific research paper, in every research field, cites relevant previous work. Each paper that is published thereby acquires a "life history", which is the record of which other work has cited it in which year. Such citation life-histories vary substantially from paper to paper; for example, some papers are cited strongly when young but then "die" after a few years, while other papers can take many years to acquire citations and then show an exponential increase. A PhD project in this area might, for example, study data from different research fields and/or different calendar decades --- there are many possibilities. This will be a very data-driven type of project, and would suit a student who, in addition to knowing well the principles of statistical modelling, is keen to develop skills in programming and data-management. The results will be of strong interest to researchers in many fields internationally, and also to research funders, university managers, etc.
- Google Scholar. ResearcherID and Depsy profiles
- potential PhD projects
- former PhD students
- unpublished or hard-to-find papers
Normal office hours (Term 1, 2019–20, starting Tuesday 1 Oct):
- Tuesdays 1030–1130
- Wednesdays 1130-1230.
Election exit pollsDavid's research work is behind the accurate exit-poll predictions that have been made at the last five UK general elections (in 2001, 2005, 2010, 2015 and 2017). For more details see exit polling explained.
Far-reaching statistical theory!What connects all these: nuclear weapons, rattlesnakes, coral, pneumonia, cancer, basketball?
Email: d dot firth at warwick dot ac dot uk
Voice: 44 (0)247 657 4855
PGP public key: df-key.txt
|Professor D Firth
Dept of Statistics
University of Warwick
Coventry, CV4 7AL
Office location: MSB 4.06, new Mathematical Sciences building.