more Rouen noir [jatp]

La Rochambelle, 25000⁺ coureuses! [39:29, 24⁰, 164th & 7th V2…]

As almost every year in the last decade, I have run the 10K in Caen for Courants de la Liberté, with 5000⁺ runners, on a new route completely in the city of Caen, partly downhill..! It did not go well (although I started in 3:44 on the first three k’s) as I ended up at a poor position (8th) in my category, which is not surprising with some runners now 8 years younger than I! (The runner next to me is the second V3.) And a fairly hot weather, especially for a Norman early morning…. Several runners fainted on the race or upon arrival and the faces of most runners showed the strain. But I first and primarily want to congratulate my mom for walking the 6⁻ km the previous evening despite serious health issues in the previous months, as well as my mother in-law who walked with her.

Infomocracy [book review]

Infomocracy is a novel by Malka Older set in a near future where most of the Earth is operating under a common elective system where each geographical unit of 100,000 people elect a local representative that runs this unit according to the party’s program and contributes to elect a Worldwide government, except for some non-democratic islets like Saudi Arabia. The whole novel revolves around the incoming election, with different parties trying to influence the outcome in their favour, some to the point of instating a dictature. Which does not sound that different from present times!, with the sligth difference that the whole process is controlled by Information, a sort of World Wide Web that seems to operate neutrally above states and parties, although the book does not elaborate on how this could be possible. The story is told through four main (and somewhat charicaturesque) characters, working for or against the elections and crossing paths along the novel. Certainly worth reading if not outstanding. (And definitely not “one of the greatest literary debuts in recent history”!)

The book is more interesting as a dystopia on electoral systems and the way the information revolution can produce a step back in democracy, with the systematisation of fake news and voters’ manipulation, where the marketing research group YouGov has become a party, than as a science-fiction (or politics-fiction) book. Indeed, it tries too hard to replicate The cyberpunk reference, William Gibson’s Neuromancer, with the same construct of interlacing threads, the same fascination for Japan, airports, luxury hotels, if not for brands, and a similar ninja-geek pair of characters. And with very little invention about the technology of the 21st Century.  (And a missed opportunity to exploit artificial intelligence themes and the prediction of outcomes when Information builds a fake vote database but does not seem to mind about Benford’s Law.) The acknowledgement section somewhat explains this imbalance, in that the author worked many years in humanitarian organisations and is currently completing a thesis at Science Po’ (Paris).

Rouen noir [jatp]

ACDC versus ABC

At the Bayes, Fiducial and Frequentist workshop last month, I discussed with the authors of this newly arXived paper, Approximate confidence distribution computing, Suzanne Thornton and Min-ge Xie. Which they abbreviate as ACC and not as ACDC. While I have discussed the notion of confidence distribution in some earlier posts, this paper aims at producing proper frequentist coverage within a likelihood-free setting. Given the proximity with our recent paper on the asymptotics of ABC, as well as with Li and Fearnhead (2016) parallel endeavour, it is difficult (for me) to spot the actual distinction between ACC and ABC given that we also achieve (asymptotically) proper coverage when the limiting ABC distribution is Gaussian, which is the case for a tolerance decreasing quickly enough to zero (in the sample size).

“Inference from the ABC posterior will always be difficult to justify within a Bayesian framework.”

Indeed the ACC setting is eerily similar to ABC apart from the potential of the generating distribution to be data dependent. (Which is fine when considering that the confidence distributions have no Bayesian motivation but are a tool to ensure proper frequentist coverage.) That it is “able to offer theoretical support for ABC” (p.5) is unclear to me, given both this data dependence and the constraints it imposes on the [sampling and algorithmic] setting. Similarly, I do not understand how the authors “are not committing the error of doubly using the data” (p.5) and why they should be concerned about it, standing outside the Bayesian framework. If the prior involves the data as in the Cauchy location example, it literally uses the data [once], followed by an ABC comparison between simulated and actual data, that uses the data [a second time].

“Rather than engaging in a pursuit to define a moving target such as [a range of posterior distributions], ACC maintains a consistently clear frequentist interpretation (…) and thereby offers a consistently cohesive interpretation of likelihood-free methods.”

The frequentist coverage guarantee comes from a bootstrap-like assumption that [with tolerance equal to zero] the distribution of the ABC/ACC/ACDC random parameter around an estimate of the parameter given the summary statistic is identical to the [frequentist] distribution of this estimate around the true parameter [given the true parameter, although this conditioning makes no sense outside a Bayesian framework]. (There must be a typo in the paper when the authors define [p.10] the estimator as minimising the derivative of the density of the summary statistic, while still calling it an MLE.) That this bootstrap-like assumption holds is established (in Theorem 1) under a CLT on this MLE and assumptions on the data-dependent proposal that connect it to the density of the summary statistic. Connection that seem to imply a data-dependence as well as a certain knowledge about this density. What I find most surprising in this derivation is the total absence of conditions or even discussion on the tolerance level which, as we have shown, is paramount to the validation or invalidation of ABC inference. It sounds like the authors of Approximate confidence distribution computing are setting ε equal to zero for those theoretical derivations. While in practice they apply rules [for choosing ε] they do not voice out, but which result in very different acceptance rates for the ACC version they oppose to an ABC version. (In all illustrations, it seems that ε=0.1, which does not make much sense.) All in all, I am thus rather skeptical about the practical implications of the paper in that it seems to achieve confidence guarantees by first assuming proper if implicit choices of summary statistics and parameter generating distribution.

Alex Honnold free solos Freeride (5.13a/7c+)

a bar for MCMC algorithms [jatp]