I'm an MRC career development award fellow (Imperial College London) and honorary lecturer (University of Bristol). I lead the multidisciplinary EpiEng (Epidemiological Engineering) group.
We work on translating ideas from engineering and statistics into better data-driven solutions to biomedical problems. Our interests sit within two themes:
Minimal modelling: minimising model complexity and maximising the information extracted from single and combined data sources
- Real-time tracking of infectious diseases from scarce data (optimal filtering and smoothing algorithms)
- Statistics for disaggregating the influence of data and prior assumptions on posterior estimates (minimum description length)
- Models for leveraging multiple data types (genetic sequences, cases, wastewater) to improve bias and variance (data fusion)
Smart modelling: succintly modelling coupled or interacting processes to devise more effective data-driven interventions
- Statistical decision theory for when to act using data across multiple scales (experimental design)
- Feedback control laws for efficiently suppressing spreading processes (reinforcement learning)
- Models of behavioural and other coupled interactions to anticipate changepoints in dynamics (networks and metapopulations)
The approach of the EpiEng group is sketched below:
![Screenshot 2024-01-23 at 18 14 42](https://private-user-images.githubusercontent.com/20108686/299057374-c463ab34-7609-4a6b-b44a-adbac2f54d1b.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.Onuf3WXEEmzdcNGnvHH7TYW6CaEQMWQLb_RQ7cmlSQI)
Example analysis: we apply sequential Bayesian algorithms (see EpiFilter at https://github.com/kpzoo/EpiFilter) to remove noise and characterise transmission rates (the time-varying smoothed reproduction number R in red) against the policies implemented in New Zealand. This is a difficult problem due to the large periods of little data, which can destabilise some predominant approaches. The use of smoothing allowed better information extraction from scarce data and yielded convincing associations between rates and policies. We also developed probabilistic indicators of when the epidemic is over (elimination probabilities Z in blue).
![](https://private-user-images.githubusercontent.com/20108686/299038077-a736da55-a4a9-4582-aa6a-cfd6ef70a3e2.jpg?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.B-Kk3S_BOPYWEcttBR2YHKImADc4jDY3XONsUui3oUE)
If interested in collaborating or applying for funding to work with us do get in touch at k.parag@imperial.ac.uk.
Keywords: Statistics, Information Theory, Control Theory, Reinforcement Learning, Phylodynamics, Bayesian Inference, Markov Processes, Epidemic Modelling, Stochastic Processes, Queueing Theory, Infectious Diseases, Time Series, Population Dynamics