Research title: Heterogeneous and latent effort modeling for non-systematic wildlife and human health data
My original interest and training was in ecology, particularly pollination ecology. However, following my undergraduate degree and a year of fieldwork, I realised that I wasn't able to answer the questions I was interested in properly without developing my quantitative skills (which were genuinely absent). A friend recommended the University of Glasgow MSc in Quantitative Methods in Biodiversity, Conservation and Epidemiology (QMBCE). I was initially reluctant as a lot of Master's courses I had looked at elsewhere seemed like their main focus was transferring money from students to universities. However, after talking more with my friend and meeting with Roman Biek, the course organiser, I was convinced. This was probably the best decision I ever made. During the MSc I went from finding statistics an opaque block to being kind of cool and eventually becoming intuitive. During my MSc thesis with Jason Matthiopoulos looking at modelling citizen science data I officially made the transition from being a pollination ecologist to someone quantitative with much broader interests (although I still love anything bee related, both academically and aesthetically).
My overarching interest is the modelling of observation processes. In ecology and epidemiology, we are rarely able to directly observe the phenomena we're interested in. Sometimes this is because it is occurring at a temporal and spatial scale beyond practical observation (at least under current funding models!). More often it's because we need to make observations measurable, so complex animal behaviours become strings of GPS locations, epidemic trajectories become aggregates of positive and negative tests across medical clinics. Traditional analytical methods and experimental design are simply not set up to deal with these problems. However, modern statistical modelling paradigms (shout out to principled Bayesian workflows) and the ubiquity of high powered computers mean that we can combine what we know of the science with what we know of the data generating process to better understand both. I see my role as trying to translate between field scientists and statisticians, and between field science and statistics.
- Describing observation problems in ecology;
- Estimating recorder effort when it isn't explicitly recorded (citizen science, apologies, this term is highly imperfect but there doesn't seem to be a better one)*;
- Statistically correcting species misclassifications (citizen science) with Otso Ovaskainen, University of Jyväskylä funded by the Eva Crane Trust.
- Exploring the relationship between multimorbidity and mortality*;
- Estimating correlations between diseases from aggregated COVID-19 mortality data;
- Combining rapid antigen testing and syndromic data to improve sensitivity and specificity in real-world COVID-19 detection.
* indicates projects put on hold due to COVID-19.
A statistical approach to tackling the misidentification crisis in bumblebee research. The Eva Crane Trust (ECTA_20200307) £7,892.92
I assist in teaching the following IBAHCM PGT courses:
- Advanced Statistics (Course Lead: Professor Daniel T Haydon);
- Spatial Ecology (Course Lead: Professor Jason Matthiopoulos);
- Introduction to Bayesian Statistics (Course Lead: Professor Jason Matthiopoulos).