Royal Statistical Society
|Dr Claus Mayer (Chair)
Biomathematics & Statistics Scotland
Aberdeen AB25 2ZD
|Dr David McLernon (Hon Secretary)
Aberdeen AB25 2ZD
|Phone: +44 (0) 1224 438652||Phone: +44 (0) 1224 437152|
Here are details about the talks planned for our next Joint Meeting with St Andrews
Wednesday, 18th of April 2018 - Joint RSS Highland group- St Andrews meeting.
Venue: Lecture Theatre C, Physics building, North Haugh.
2.00pm -2.05pm Welcome
2.05pm – 2.55pm Marta Blangiardo (Imperial)
2.55pm – 3.30pm Coffee Break
3.30pm - 4.20pm Thordis L. Thorarinsdottir (Norwegian Computing Centre, Oslo)
Marta Blangiardo is a senior lecturer in Biostatistics at the Department of Epidemiology and Biostatistics, Imperial College London (UK) affiliated with the MRC-PHE Centre for Environment and Health (http://www.environment-health.ac.uk/). Her main research interest is the development of Bayesian hierarchical models, especially in relation to the use of multiple data sources, with application in a wide range of areas, including genetics, environmental science, and epidemiology. Recently she has been working on spatio-temporal epidemiological surveillance for chronic diseases, confounder adjustment in ecological studies of environmental epidemiology and models for multi-pollutant exposure in a time series framework.
A data integration approach for improving inference in area-referenced environmental health studies (joint work with Monica Pirani, Alexina Mason, Anna Hansell, and Sylvia Richardson)
Abstract: Study designs where data are aggregated into geographical areas are extremely popular in environmental epidemiology. These studies are commonly based on administrative databases and, providing a complete spatial coverage, are particularly appealing to make inference on the entire population. The ecological nature of these studies, however, does not allow the direct inclusion of individual-level risk factor data. In the presence of unmeasured potential confounding factors, risk effect estimates are prone to bias. Here, we show how to improve inference drawn from area-referenced environmental health-effect studies, proposing a Bayesian approach that augments measured area-level covariates with an ecological propensity score estimated upon individual-level data from sample surveys. This scalar index acts as a proxy for the unmeasured ecological confounders and represents a useful tool for overcoming the problem due to the incomplete spatial coverage of the individual-level data. In contrast to the main literature on propensity score for confounding adjustment where the exposure of interest is confined to a binary domain, we generalize its use to cope with ecological studies characterised by a continuous exposure. The approach is illustrated using simulated examples and a real application investigating the risk of lung cancer mortality associated to nitrogen dioxide in England (UK).
Thordis L. Thorarinsdottir is a statistician working within the fields of spatial and space-time statistics, Bayesian methods and forecasting. She is a Chief Research Scientist in the Statistical Analysis, Machine Learning, and Image Analysis (SAMBA) group at the Norwegian Computing Center in Oslo, Norway. Her current work focuses on environmental applications to solve problems on uncertainty quantification, probabilistic prediction and model evaluation in collaborations with climate ecologists, atmospheric scientists and hydrologists.
Title: Does Bayes beat squinting? Bayesian modelling of cluster point process models
Abstract: Point process data arises naturally in various fields of science such as biology, ecology, epidemiology, and environmental sciences. However, the point process modelling framework is very involved and inference can often only be performed approximately and with great care. At the same time, a great number of different models are available where the subtle differences between the individual models can be hard to detect. In this talk, we discuss to which extend Bayesian modelling approaches may be applied to the class of cluster process models. Cluster point processes have the following general structure. There is a point process of cluster centers and to each cluster center is associated a random number of points forming a subsidiary process, where the points in the subsidiary process are distributed about the cluster center in some specific way. A model for a cluster point process thus consists of three components; a component describing the cluster center process, a component describing the cluster sizes, and a component describing the distribution of the subsidiary points around the cluster center, the dispersion process. We consider how Bayesian approaches may be used to perform inferences for all three components, a feat which is often not possible using other inference methods