Research

My research develops statistical methods for dependence and dynamics in complex data, and applies them to problems in finance, insurance, energy, and beyond. A few connected lines run through the work.

Area 01

Local dependence and nonlinear modelling

Much of my work builds a framework for modelling local features of multivariate distributions through local Gaussian approximations — the idea that complicated global structure becomes tractable when viewed locally. It has produced density estimators, measures of local and partial dependence, tools for conditional association and classification, and tests for conditional independence, all able to capture the nonlinear and asymmetric dependence that global, linear methods miss. The framework is gathered in a monograph (Academic Press) and made usable through the lg package on CRAN.

local Gaussian framework · book · R package

Area 02

Statistical modelling of time series: regimes, volatility & extremes

I am increasingly drawn to statistical models for time series whose behaviour shifts over time and where rare, extreme events carry most of the risk. Recent work fits hidden semi-Markov models to heavy-tailed, temporally clustered insurance claims, and studies local, nonlinear lead–lag structure in financial markets.

regimes · heavy tails · volatility

Area 03

Risk and decision-making under uncertainty

A third line uses statistical modelling to support decisions when the future is uncertain. It spans portfolio allocation under asymmetric and nonlinear dependence — where diversification tends to fail exactly when it is needed most — insurance risk under a changing climate, and the planning of renewable energy, from the geographical allocation of offshore wind to seasonal renewable portfolios. The common thread is to quantify risk with models that reflect the data, then turn the results into practical choices.

Climate Futures (SFI) · Finance Market Fund

Area 04

Applied statistics across disciplines

I also value cross-disciplinary collaboration. This includes expository work that frames statistical dependence for a broad audience, machine learning applied to law and economics (predicting patent litigation), causal analysis of how shareholder activism shapes green innovation, and Bayesian graded-response models for fair, comparable grading across student cohorts.

collaborative · cross-disciplinary