Modelling changepoints in a Bayesian way can be an extremely elegant and computationally feasible way of doing inference in an on-line way. I am aiming to extend this inherent time series method into a spatio-temporal context admitting non-stationary behaviour in space and time while retaining tractability enabling scalable inference. Applying an algorithm like this to London's air pollution sensor network can reveal whether changes in congestion charges achieve the desired fall in the level of pollutants.
Before joining OxWaSP, I had a very strong interest in time series methods. This ranged from applied work on European temperature trends as part of my Bachelor thesis (journals.library.tudelft.nl/index.php/sure/article/view/1074) to investigating the validity of post-model selection inference in my Master thesis (https://graduationtheses.library.maastrichtuniversity.nl/ThesisDetails?id=guid:f8d9aefc-5fa7-4c68-a9a6-a2434e8dd778)
Office: C1.02, Zeeman building