Dr Oliver Stoner
- Lecturer (Statistics)
I am an early career Lecturer in Statistics. I am motivated by application-driven statistical modelling problems and have a keen interest in developing general frameworks and bespoke models using Bayesian hierarchical methods.
Prior to joining Glasgow, I was a Research Fellow for 2 years at the University of Exeter, where I also obtained my PhD in 2019. My thesis is entitled Bayesian Hierarchical Modelling Frameworks for Flawed Data in Environment and Health.
The two main projects I am leading are an impact-oriented collaboration with the World Health Organization and a more methodologically-driven research programme in correcting delayed reporting of infectious diseases.
Household Air Pollution
Around 2.4 billion people mainly use polluting fuels for cooking, hampering their socio-economic prospects and exposing them to dangerous levels of household air pollution, attributed by the World Health Organization to 3.2 million deaths per year in 2020. I have worked with the WHO to develop a bespoke Bayesian hierarchical model for estimating the populations using 6 different fuels for cooking, called the Global Household Energy Model. Estimates from this model serve as the primary basis for tracking access to clean cooking as part of UN Sustainable Development Goal 7 (SDG 7.1.2), and are a key input to WHO burden of disease calculations for household air pollution (SDG 3.9.1).
Our paper presenting these estimates was published Nature Communications in 2021 and has been gaining traction. I have also contributed as a named author to the 2020, 2021 and 2022 editions of the internationally recognised report Tracking SDG7: The Energy Progress Report.
Correcting Delayed Reporting of Infectious Diseases
Surveillance of infectious disease cases and/or deaths is often subject to delayed reporting, where complete information is not immediately available to decision-makers. Statistical models can learn the systematic and random structures in the reporting delay, to predict current case/death counts based on incomplete data.
The best introduction to this work I can offer is an article we prepared for Significance Magazine early on in the COVID-19 pandemic: A more transparent way to estimate and report daily Covid-19 deaths, aimed at a general statistical audience.
During my PhD we developed a new multivariate hierarchical framework for correcting delayed reporting. This was published in Biometrics just before COVID-19 started appearing in our news feeds, and as a postdoc I sought to demonstrate this method as a compelling operational tool for pandemic decision-making. Our method was used by researchers at Cambridge MRC Biostatistics Unit as part of their efforts to nowcast COVID-19. Their estimates regularly informed the Scientific Pandemic Influenza Group on Modelling (SPI-M) and therefore fed into UK Government decision-making.
Our work on delayed reporting followed our work on a new method for correcting under-reporting, published in the Journal of the American Statistical Association in 2019.
- Statistics & Data Analytics
- Modelling in space and time
- Bayesian modelling and inference
- Computational statistics
- Environmental, ecological sciences and sustainability
- Biostatistics, epidemiology, and health applications
- Social and urban studies
EPSRC Impact Accelerator Account (Ongoing, PI)
Funding for half my time, initially for 9 months, to work with the World Health Organization to develop the first worldwide estimates of the number of people using polluting fuels for household heating. Our aim is to inform new policy efforts to reduce polluting heating, which is responsible for significant health impacts, unsustainable harvesting of wood fuels, and climate-damaging emissions.
UK Department for Business, Energy and Industrial Strategy Regulators' Pioneer Fund (2022, CoI)
Funding to improve monitoring of harmful algal blooms, through improvements to statistical forecasting models and trialling of new quantitative polymerase chain reaction (qPCR) measurement methods.
IIB Open Innovation Platform: Collaboration Fund (2021, CoI)
Funding to develop advanced prediction models for harmful algal blooms and impacts on the shellfish industry. Paper submitted to Nature Food in June 2021.
ESRC Impact Accelerator Sub-Award: Knowledge Exchange Fellowship (2021, PI)
Funding to support a collaborative paper with the WHO on modelling polluting household energy.
I have also been PI on two research contracts between the WHO and the University of Exeter, to support annual updates to Sustainable Development Goal 7 and to contribute to associated reporting.
I am always interested to supervise motivated and capable people to study a PhD in Statistics, either as lead-supervisor or part of a supervisory team. My main area of expertise that I can contribute to supervision is on Bayesian hierarchical modelling methods and software.
Please contact me by email to discuss project ideas and opportunities for fully-funded studentships.
- Halliday, Alba
Advanced prediction models for infectious diseases
- Holland, Catherine
Bayesian approaches to compositional data with structural zeros
- Zhu, Qiangqiang
Spatio-temporal modelling of population-level disease risk when the populations at risk have partially unknown spatial locations
I am currently supervising PhD student Alba Halliday, who is working on developing operational models for spatio-temporal disease prediction.
In the 2022/23 academic year, I will be teaching Advanced Data Analysis and supervising MSc projects.
I am currently a 2nd-year student on the Postgraduate Certificate in Academic Practice programme.