Many candidates have done student projects on casualty loss regression, following the templates on the NEAS web site. This student project is particularly good, and shows how the analysis can be done. Note that even a slight increase in the stochasticity (from 2% to 3%) makes the results less clear, making actual regression studies difficult.