Bayesian Inference for stochastic processes

BioSS has made significant contributions to development of inference methods and their application to a range of important societal challenges

Models + inference make sense of data

  • Real world data typically provides a partial and unreliable picture
  • Inference for dynamic processes enables estimation  - from real world data - of critical features of systems not amenable to direct measurement.


We develop novel - and apply existing -tools to tackle these inference challenges:

  • computational speed – we dev. efficient exact & approx. methods
  • integrating genetic data with more standard observations
  • accounting for heterogeneity in space and time & between individuals
  • application to large-scale problems and models of social-systems


  • Infectious disease in livestock & wildlife e.g. bovine TB, African Swine Fever, Avian influenza
  • Spread of invasive species – UK invasive vascular plants
  • Forest pests – Emerald Ash Borer, Great Spruce bark beetle
  • COVID-19
  • Environmental transmission
  • Host genetic effects in disease
  • Population dynamics
  • Behaviour in social systems

Stochastic process models

  • Ideal where chance events play a critical role e.g. disease transmission
  • Widely used to model chemical, ecological and epidemiological populations and social systems
  • Require parameters – in example shown describing time duration of disease states and infectivity
  • Use data to estimate parameters and missing events – in the example shown only see individuals that test positive but not when they become infected
  • Also infer model structure – e.g. transmission routes

Example: disease dynamics

  • Observations limited by diagnostic tests for disease

Observations limited by diagnostic tests for disease