Danny WilliamsChicago Crime
My most recent project was a group project as part of the first term of my CDT. It involved modelling different aspects of a large data set detailing crimes in Chicago. My input to this project was to use logistic regression methods to classify whether an arrest would be successful or not, given other covariates in the crime set; detailing the location, the area, the type of crime type, amongst others.
The package that was created for this project can be found as a Github repository here.
Downscaling Extremes of Precipitation in the South West
For my Masters dissertation project, I used extreme value theory to model the extremes of precipitation across the South West of the UK. Using extreme value theory to model low resolution gridded model output, and observations at a local level, a separate downscaling model was able to accurately predict for a new area where local observations are sparse.
The final dissertation can be found here.
Post-Processing Ordinal Cloud Cover Forecasts
During a summer project at the University of Bristol, I developed methodology that extended the polynomial odds logistic regression approach that is currently used to post-process weather forecasts for total cloud cover from an ensemble forecast model. This used Bayesian methods, using MCMC to increase flexibility.