Subjective Bayes at 48y

politics coming [too close to] statistics [or the reverse]

On 30 April, David Spiegelhalter wrote an opinion column in The Guardian, Coronavirus deaths: how does Britain compare with other countries?, where he pointed out the difficulty, even “for a bean-counting statistician to count deaths”, as the reported figures are undercounts, and stated that “many feel that excess deaths give a truer picture of the impact of an epidemic“. Which, on the side, I indeed believe is a more objective material, as also reported by INSEE and INED in France.

“…my cold, statistical approach is to wait until the end of the year, and the years after that, when we can count the excess deaths. Until then, this grim contest won’t produce any league tables we can rely on.” D. Spiegelhalter

My understanding of the tribune is that the quick accumulation of raw numbers, even for deaths, and their use in the comparison of procedures and countries is not helping in understanding the impacts of policies and actions-reactions from a week ago. Starting with the delays in reporting death certificates, as again illustrated by the ten day lag in the INSEE reports. And accounting for covariates such as population density, economic and health indicators. (The graph below for instance relies on deaths so far attributed to COVID-19 rather than on excess deaths, while these attributions depend on the country policy and its official statistics capacities.)

“Polite request to PM and others: please stop using my Guardian article to claim we cannot make any international comparisons yet. I refer only to detailed league tables—of course we should now use other countries to try and learn why our numbers are high.” D. Spiegelhalter

However, when on 6 May Boris Johnson used this Guardian article during prime minister’s questions in the UK Parliement, to defuse a question from the Labour leader, Keir Starmer, David Spiegelhalter reacted with the above tweet, which is indeed that even with poor and undercounted data the total number of cases is much worse than predicted by the earlier models and deadlier than in neighbouring countries. Anyway, three other fellow statisticians, Phil Brown, Jim Smith (Warwick), and Henry Wynn, also reacted to David’s tribune by complaining at the lack of statistical modelling behind it and the fatalistic message it carries, advocating for model based decision-making, which would be fine if the data was not so unreliable… or if the proposed models were equipped with uncertainty bumpers accounting for misspecification and erroneous data.

head position at Warwick stats

The Department of Statistics at Warwick seeks a new head to continue to develop and advance the quality of its education and research. The successful candidate will be appointed as a professor on an indefinite basis and will have a strong research and leadership profile. The appointment as Head of Department will be for three years in the first instance, with an option to extend. The next Head will work with this large and diverse community of academics and students, and support collaboration with the wider University. They will represent the Department to public and private audiences, nationally and internationally, and develop networks to promote the work of the Department. The deadline for applicants is 28 September 2018.

 

Bayesian Decision Analysis

In 1989, I wrote my very first book review, which appeared in JASA, about “Decision Analysis: A Bayesian Approach“, by J.Q. Smith, professor at the University of Warwick. I have now received from the Short Book Reviews section of the International Statistical Review another book of Jim Smith to review, namely Bayesian Decision Analysis. Before getting on discussing the current book, let me acknowledge that my review of 1989 was far too rash and critical! While acknowledging that the 1989 book developed “concepts not usually dealt with in Bayesian classics”, I bemoaned the lack of connections with classical Bayesian decision theory, as exemplified by Berger (1985) and missing entries about noninformative priors, prior construction, bounded loss shortcomings, &tc. While I remain attached to the approach adopted in Jim [Berger]’s book,  I now see much more clearly the point made in Jim [Smith]’s 1989 book and thus regret a posteriori the tone of this review as une erreur de jeunesse…