Title: Complex models, time pressure, and COVID
Authors: Thomas House - University of Manchester (United Kingdom) [presenting]
Abstract: The focus is on some of the issues that arose while providing modelling and statistical support to the UK Government during the current pandemic. Given the fast-changing nature of the pandemic, the exact content is subject to change; however, questions include (1) Use of parametric versus non-parametric versus mechanistic models; (2) When to use asymptotic approximations (e.g. Laplace) that may be optimistic about uncertainty but allow for much faster calculations; (3) When to combine multiple analyses sources formally and when to run separate analyses and compare using expert judgement. These are, of course, questions of general interest in computational statistics.