On 28
th September,
the RSS Merseyside Local Group hosted our third event of the year, a meeting on the topic of Bayesian Real-Time Modelling. The event was hosted at the University of Liverpool with two invited external speakers. A recording of the event is available to watch back on the
RSS Merseyside YouTube channel and has received 40 views in its first three weeks online.
First, we welcomed Prof. Darren Wilkinson, Chair of Statistics at Durham University and a Fellow of the Alan Turing Institute. He began his talk on “Streaming data modelling for real-time monitoring and forecasting” by explaining the nature of streaming data and its implications for functional programming. He then highlighted the connection between streaming data and state-space modelling by presenting the Flood-PREPARED project, which uses Bayesian inference to predict short-term flood risks from rainfall data. Applying the model to Newcastle, Prof. Wilkinson showed how spatiotemporal forecasts could be generated in near real-time, informing decisions about emergency service availability.
We then welcomed Prof. Peter Neal, Professor of Statistics at the University of Nottingham. Prof. Neal introduced “Statistical inference for emerging diseases inspired by Covid-19” by recapping the SIR epidemic model and a likelihood derivation for disease recovery times. He then showed how this likelihood becomes more efficient when approximating a birth-death process, no longer relying on disease infection times and speeding up calculations by an order of magnitude when applied to a demo smallpox outbreak dataset. Prof. Neal then scaled this approach up to model the early COVID-19 pandemic in Europe, allowing efficient estimation of parameters associated with control measures.
Talks were followed by audience questions, which covered topics such as which metrics are appropriate to validate these sophisticated models and how one can potentially identify the effects of biases in input data. Overall, the event showcased some exciting developments in Bayesian statistical frameworks that can help us give timely responses to emerging national issues through real-time analysis, as well as highlighting remaining computational challenges and the need for equally innovative smoothing and sampling algorithms.
The RSS Merseyside Local Group plans to host further in-person events in December featuring talks on Statistics in Cybersecurity and in February featuring talks on Neglected Tropical Diseases. Our AGM will take place in December and the date will be announced shortly.