My impression, admittedly superficial, from the ongoing debate on the severity of the Covid-19 pandemic is that statisticians with a professional interest in the distribution of extrema (“extreme value distributions”) understand the downside risks well and, as a result, support the most drastic measures to curb those risks.
Here’s a piece from Monday, April 13, from The Independent:
Dr Holger Rootzen, one of the earliest detractors of Sweden’s policy and a professor of mathematical statistics at the University of Chalmers, says that Sweden will soon be overwhelmed unless more action in taken.
Prof. Rootzén is a co-author of Extremes and Related Properties of Random Sequences and Processes, first published as early as in 1983. The underlying mathematics isn’t all that new but the general public has started paying attention only recently, and reluctantly.
Not to travel too far from Gothenburg, where Rootzén is based, here’s an extract from the description of an advanced statistics course offered by the University of Oslo, “Extreme value statistics and large deviations”:
This course offers an introduction to basic concepts about extreme value theory and the theory of large deviations, and the estimation of tails of probability distribution for actuarial reserves.
On completing the course, the student would be expected to have learned, among other things, “about… distributions for the maxima of random variables” and “how to estimate the probability of extremely rare events and rate of exceedance.” She would also “become familiar with applications of the theory of extreme values such the estimation of distributions of, e.g., flood, earthquakes, big insurance losses, equity risk, etc.”
I would add recurring epidemics. Covid-19? There’s not much data for reliable inference but this kind of training should at least equip statisticians with the optics to spot others’ errors.