Statistical methodology for handling missing data when modelling extremes
Project ID: 2531ad1585
(You will need this ID for your application)
Research Theme: Mathematical Sciences
UCL Lead department: Statistical Science
Lead Supervisor: Robin Mitra
Project Summary:
Research importance:
Understanding extreme events is crucial in many settings, from the environment to finance. Although these are generally rare occurrences, they often have a big impact. When we talk about extremes, we mean values that occur in the tails of a distribution, and statistical techniques tailored to the bulk generally perform poorly in these regions. Extreme value analysis provides a mathematically-motivated framework for the statistical modelling of extreme events, able to cover a range of problems. However, considerations around missing data are often neglected in extreme value modelling; the aim of this project is to develop novel statistical techniques for extremes that are robust to such missingness.
Supervisory team:
The project will be supervised by Robin Mitra and Emma Simpson, both from the Department of Statistical Science at UCL. Their research expertise is on missing data and extreme value analysis, respectively, and this project aims to combine these two important and highly topical areas.
What you will be doing:
The focus of the project will be on developing new statistical methodology for missing data and extremes. We will begin by investigating the issues that arise with different definitions of “extreme events” and different missing data processes. We will then aim to develop new statistical methods to handle these problems - this could be tackled from frequentist or Bayesian perspectives. The day-to-day research will involve developing the mathematics to advance our understanding on the theory underlying these approaches, computational implementation of new ideas to test their performance, and working on ways the methodology could be deployed in relevant applied settings.
Who we are looking for:
We do not expect any previous experience in missing data or extremes specifically, but a strong background in statistics and/or mathematics more generally is important. Experience of programming, particularly in R, would also be helpful.