Models and Uncertainty in the time of COVID-19
Professor Sally Cripps
School of Mathematics and Statistics
Director, ARC Centre for Data Analytics for Resources and Environments (DARE)
University of Sydney
6:00 ᴘᴍ on Tuesday 13th October 2020
Kim E. Beazley Lecture Theatre
Murdoch University, Western Australia
Models for forecasting and inference have been influential in shaping decision-making in the COVID-19 pandemic. However, there is concern that predictions and inference from these models may be misleading. Misleading in two senses; first in the accuracy of any given model, where this accuracy is measured with respect to the entire predictive distribution, we call this prediction uncertainty, and second, in the variability of conclusion drawn from using different models, which we call inference uncertainty. Both sources of uncertainty are important, and decision makers who are unaware of, or ignore, either source of uncertainty are underestimating the risk attached to any decision based on that model. In this talk we will examine the prediction uncertainty of several forecasting models used in the USA in the early stages of the pandemic, as well as the inference uncertainty surrounding the effectiveness of various non-pharmaceutical interventions (NPIs using different models produced by Imperial College. We also present a method for combining models to produce more accurate posterior predictive distributions. We conclude by noting that our role as statisticians is to quantify and report all sources of uncertainty; we must be transparent about what we do not know. Failure to communicate this uncertainty will ultimately undermine the public's trust in the value of decisions based on statistical modelling.
About the Speaker
Sally Cripps is a Professor of Mathematics and Statistics and Director of the ARC Centre in Data Analytics for Resources and Environments (DARE Centre), at the University of Sydney. Sally’s research focus is the development of new and novel probabilistic models which are motivated by the need to solve an applied problem with the potential for impact. She has particular expertise in the use of mixture models for complex phenomena, modelling longitudinal data, nonparametric regression, the spectral analysis of time series, and the construction of transition kernels in MCMC schemes that efficiently explore posterior distributions of interest. Sally is also Chair of the International Society for Bayesian Analysis’ section, Bayesian Education and Research in Practice.
Refreshments and Dinner
Members and guests are invited to mingle over wine and cheese from 5:30 ᴘᴍ onwards in the Kim E. Beazley Lecture Theatre. Following the meeting you are invited to dine with fellow attendees at The Kardinya Tavern (17 South St, Kardinya); please RSVP to Brenton Clarke B.Clarke@murdoch.edu.au by 11 am Tuesday 13th October.
Parking is free at the Murdoch University campus after 5:00 ᴘᴍ in the red and green zones (avoid the reserved and service bays). We suggest parking in Car Park 3 (searchable in Google Maps) that can be entered from campus entrance B or C on South St; those parking here should walk south-west towards Bush court then up the stairs when seen.
The Kim E. Beazley Lecture theatre entrance with glass doors is on the south side of Bush Court. The approximate location is indicated by the yellow start on the map below, or use this Google Maps link.
Alternatively, view online with zoom
For those that can't make it physically, please register on this page for Zoom connection details.
Visitors are welcome to attend this branch meeting.
For further information please contact the Branch Secretary, Rick Tankard, Murdoch University.
He can be reached by email at email@example.com or by phone at (08) 9360 2820.