Drop me an email if you are interested in any of the projects below. Alternatively, feel free to pitch your own project! Contact me at s.liverani@qmul.ac.uk
- Bayesian spatial modelling for biodiversity (collaboration with Royal Botanical Gardens, Kew)
Many models assume that observations are obtained independently of each other.
However, distance between observations can be a source of correlation, which needs to be
accounted for in any model. For example, pollution has a spatial smooth pattern and
measurements close in space are likely to be very similar. Spatial models therefore have to
take into account the spatial autocorrelation in datasets in order to separate the general
trend (usually depending on some covariates) from the purely spatial random variation.
This project will focus on developing and applying Bayesian spatial and spatio–temporal
modelling techniques to enhance our understanding of the association between, and
predict, plant species that are at risk of extinction and areas in need of protection in the face
of climate change, changing land use (especially agriculture) and pollution. The pollutants of
interest are nitrogen and phosphate–based fertilizers. We will leverage spatial distribution
data for the entire British flora, studying changing trends in distribution and land use over
60 years. We will also include data for a range of different measures of genomic diversity
(e.g. genome size and polyploidy) together with climate and soil data to uncover the role of
biological and abiotic factors in predicting species at risk of extinction, and landscapes at
increased risk of biodiversity loss under differing land use and climate change scenarios. The
research will be undertaken in collaboration with Dr Ilia Leitch, Senior Research Leader at
the Royal Botanic Gardens, Kew. One statistical challenge that arises is that the data are available at different resolutions, and advanced methods are required to model misaligned spatial and spatio–temporal data. We will leverage recent work by Dr Silvia Liverani on Bayesian methods for misaligned areal
data, and extend them to meet the needs of this research challenge in the study and
understanding of biodiversity. - Bayesian modelling for misaligned spatio-temporal data
This project will focus on Bayesian spatial modelling techniques to examine the association between measures of population health (such as hospitalisations and mortality) and socio-economic factors. This will be accomplished by developing spatial and spatio-temporal models which can allow for spatially varying coefficients, spatially adaptive priors, and spatial quantile regression. - Clustering different types of data
Many different kinds of data can be used with algorithms of clustering. The data can be like binary data, categorical and interval-based data, or a mixture of these types of data. However, when it is a mixture of different types of data, it is not clear how this affects the results. This PhD project will study this for Dirichlet process mixture models, using and extending the R package PReMiuM.