People and Planet Group
The Statistics for People and Planet group is an informal, broad-church research group for academics and PhD students interested in applications of Statistics to the Social, Environmental and Ecological Sciences. We meet fortnightly, and our members give a short talk about their research, recent papers or anything else of interest.
New paper: Modelling particle number size distribution
Our recent paper presents a new model for number particle size distributions (NPSD) that covers fine and ultrafine particles, which are known to be detrimental to health.
We use a functional data approach to cope with the high-dimensionality and high correlation characteristics of the data. With the proposed model, we found six main sources of particulate matter. You can try our new approach using our developed SAfda R package.
New paper: Forecasting intraday particle number size distribution
This paper provides a new method to forecast non-stationary and high-dimetional number particle size distributions data (NPSD). Due to the complex dependence structure, we propose to use the functional data approach – each observation is assumed to be a continuous curve. Forecast can obtain with very low computational cost (seconds on standard computers).
New paper: Fast return level estimates for flood insurance
Existing methods for quantifying the distribution of high total yearly insurance claims due to flooding can take days or weeks to run. They involve simulating thousands of years of data, and for each year, each individual claim for each hypothetical flood event is simulated. We develop a new concentration inequalty which forms the basis for a novel methodology that simulates a relatively tight upper bound on the total loss for each year, allowing estimation of quantities of interest, such as return levels, several orders of magnitude more quickly than previously, and enabling hitherto infeasible sensitivity studies.