Dependence modelling using truncated vine copulas with applications

About the Project

Primary Supervisor: Dr. Aristidis K. Nikoloulopoulos

Multivariate response data abound in many applications including insurance, risk management, finance, psychometrics, health and environmental sciences. Data from these application areas have different dependence structures. While a multivariate distribution fully encodes this dependence, the tractable families used in practice often impose restrictive marginal or dependence structures. Copula functions alleviate these constraints by separating the margins from the dependence structure. Although classical copulas are naturally suited to low-dimensional settings, vine copulas extend the framework to high dimensions. We have shown that a vine copula displays (tail) dependence in all bivariate margins provided that the pair-copulas in the first level possess (tail) dependence; higher-level pair-copulas may be independence copulas without loss of overall (tail) dependence. This insight justifies truncating the vine after the first level, creating a parsimonious model that retains the essential dependence structure. In this project, we will make use of truncated vine copulas with both observed and latent variables in the aforementioned application areas.

Entry Requirements

The entry requirements are either a 1st in your Bachelor's degree or a Master's in Mathematics, Statistics, or Actuarial Science.

Mode of Study

Full or Part time

Start Date

1 October 2026