Research


Find me on:

  • Google Scholar
  • GitHub
  • UCD Research
  • ResearchGate
  • LinkedIn
  • Orcid
  • Twitter

For recent talks see here.

Palaeoclimate reconstruction

My main field of research for the last 10 years has been Bayesian reconstructions of palaeoclimate from proxy data such as pollen, foraminifera, and stable isotopes. Projects under this topic include: Multivariate misaligned time series for ice core and pollen records, Estimating worldwide rates of relative sea level change, and Identifying frequency behaviour of palaeoclimate records using Bayesian variable selection methods. This research involves a substantial computing element, and combines ideas in stochastic processes, state-space modelling, and long-tailed probability distributions. Through this work we have developed the R packages Bchron and Bclim.

Text mining

As principal investigator of the Enterprise Ireland Innovation Partnership Automatic Dynamic Product Classification I work on statistical and machine learning methods for classification of text data. We are currently applying these approaches to very large data sets provided by a Dublin-based company concerning product classifications for online shopping websites.

Extremes

As principal investigator of the Environmental Protection Agency funded project Extreme Events in Calibrated Climate Models: Impacts for Ireland I work on adapting extreme value distributions to spatio-temporal, multivariate scenarios. We are currently applying these methods to data supplied by Met Eireann, and also to wave buoy from Ireland's west coast.

Quantitative ecology

My interest in quantitative ecology concerns Bayesian compositional modelling of data from stable isotope experiments on ecosystems. Through this research we have developed the widely-used R computer package SIAR and its successors simmr and MixSIAR.

Bioinformatics

As part of the SFI/SBI/IRCSET-funded Bioinformatics PhD scheme, I am involved in developing model-based machine learning methods for biomarker discovery. The methods enable better prediction of the occurrence and severity of prostate cancers.