revisiting marginalisation paradoxes [Bayesian reads #1]

As a reading suggestion for my (last) OxWaSP Bayesian course at Oxford, I included the classic 1973 Marginalisation paradoxes by Phil Dawid, Mervyn Stone [whom I met when visiting UCL in 1992 since he was sharing an office with my friend Costas Goutis], and Jim Zidek. Paper that also appears in my (recent) slides as an exercise. And has been discussed many times on this  ‘Og.

Reading the paper in the train to Oxford was quite pleasant, with a few discoveries like an interesting pike at Fraser’s structural (crypto-fiducial?!) distributions that “do not need Bayesian improper priors to fall into the same paradoxes”. And a most fascinating if surprising inclusion of the Box-Müller random generator in an argument, something of a precursor to perfect sampling (?). And a clear declaration that (right-Haar) invariant priors are at the source of the resolution of the paradox. With a much less clear notion of “un-Bayesian priors” as those leading to a paradox. Especially when the authors exhibit a red herring where the paradox cannot disappear, no matter what the prior is. Rich discussion (with none of the current 400 word length constraint), including the suggestion of neutral points, namely those that do identify a posterior, whatever that means. Funny conclusion, as well:

“In Stone and Dawid’s Biometrika paper, B1 promised never to use improper priors again. That resolution was short-lived and let us hope that these two blinkered Bayesians will find a way out of their present confusion and make another comeback.” D.J. Bartholomew (LSE)

and another

“An eminent Oxford statistician with decidedly mathematical inclinations once remarked to me that he was in favour of Bayesian theory because it made statisticians learn about Haar measure.” A.D. McLaren (Glasgow)

and yet another

“The fundamentals of statistical inference lie beneath a sea of mathematics and scientific opinion that is polluted with red herrings, not all spawned by Bayesians of course.” G.N. Wilkinson (Rothamsted Station)

Lindley’s discussion is more serious if not unkind. Dennis Lindley essentially follows the lead of the authors to conclude that “improper priors must go”. To the point of retracting what was written in his book! Although concluding about the consequences for standard statistics, since they allow for admissible procedures that are associated with improper priors. If the later must go, the former must go as well!!! (A bit of sophistry involved in this argument…) Efron’s point is more constructive in this regard since he recalls the dangers of using proper priors with huge variance. And the little hope one can hold about having a prior that is uninformative in every dimension. (A point much more blatantly expressed by Dickey mocking “magic unique prior distributions”.) And Dempster points out even more clearly that the fundamental difficulty with these paradoxes is that the prior marginal does not exist. Don Fraser may be the most brutal discussant of all, stating that the paradoxes are not new and that “the conclusions are erroneous or unfounded”. Also complaining about Lindley’s review of his book [suggesting prior integration could save the day] in Biometrika, where he was not allowed a rejoinder. It reflects on the then intense opposition between Bayesians and fiducialist Fisherians. (Funny enough, given the place of these marginalisation paradoxes in his book, I was mistakenly convinced that Jaynes was one of the discussants of this historical paper. He is mentioned in the reply by the authors.)

bootstrap in Nature

A news item in the latest issue of Nature I received about Brad Efron winning the “Nobel Prize of Statistics” this year. The bootstrap is certainly an invention worth the recognition, not to mention Efron’s contribution to empirical Bayes analysis,, even though I remain overall reserved about the very notion of a Nobel prize in any field… With an appropriate XXL quote, who called the bootstrap method the ‘best statistical pain reliever ever produced’!

double yolk priors [a reply from the authors]

[Here is an email I received from Subhadeep Mukhopadhyay, one of the authors of the paper I discussed yesterday.}

Thank for discussing our work. Let me clarify the technical point that you raised:

– The difference between Leg_j(u)_j and T_j=Leg_j(G(θ)). One is orthonormal polyn of L₂[0,1] and the other one is L₂[G]. The second one is poly of rank-transform G(θ).

– As you correctly pointed out there is a danger in directly approximating the ratio. We work on it after taking the quantile transform: evaluate the ratio at g⁻¹(θ), which is the d(u;G,F) over unit interval. Now, this new transformed function is a proper density.

-Thus the ratio now becomes d(G(θ)) which can be expended into (NOT in Leg-basis) in , in eq (2.2), as it lives in the Hilbert space L₂(G)

– For your last point on Step 2 of our algo, we can also use the simple integrate command.

-Unlike traditional prior-data conflict here we attempted to answer three questions in one-shot: (i) How compatible is the pre-selected g with the given data? (ii) In the event of a conflict, can we also inform the user on the nature of misfit–finer structure that was a priori unanticipated? (iii) Finally, we would like to provide a simple, yet formal guideline for upgrading (repairing) the starting g.

Hopefully, this will clear the air. But thanks for reading the paper so carefully. Appreciate it.

double yolk priors

“To develop a “defendable and defensible” Bayesian learning model, we have to go beyond blindly ‘turning the crank’ based on a “go-as-you-like” [approximate guess] prior. A lackluster attitude towards prior modeling could lead to disastrous inference, impacting various fields from clinical drug development to presidential election forecasts. The real questions are: How can we uncover the blind spots of the conventional wisdom-based prior? How can we develop the science of prior model-building that combines both data and science [DS-prior] in a testable manner – a double-yolk Bayesian egg?”

I came through R bloggers on this presentation of a paper by Subhadeep Mukhopadhyay and Douglas Fletcher, Bayesian modelling via goodness of fit, that aims at solving all existing problems with classical Bayesian solutions, apparently! (With also apparently no awareness of David Spiegelhalter’s take on the matter.) As illustrated by both quotes, above and below:

“The two key issues of modern Bayesian statistics are: (i) establishing principled approach for distilling statistical prior that is consistent with the given data from an initial believable scientific prior; and (ii) development of a Bayes-frequentist consolidated data analysis work ow that is more effective than either of the two separately.”

(I wonder who else in this Universe would characterise “modern Bayesian statistics” in such a non-Bayesian way! And love the notion of distillation applied to priors!) The setup is actually one of empirical Bayes inference where repeated values of the parameter θ drawn from the prior are behind independent observations. Which is not the usual framework for a statistical analysis, where a single value of the parameter is supposed to hide behind the data, but most convenient for frequency based arguments behind empirical Bayes methods (which is the case here). The paper adopts a far-from-modern discourse on the “truth” of “the” prior… (Which is always conjugate in that Universe!) Instead of recognising the relativity of a statistical analysis based on a given prior.

When I tried to read the paper any further, I hit a wall as I could not understand the principle described therein. And how it “consolidates Bayes and frequentist, parametric and nonparametric, subjective and objective, quantile and information-theoretic philosophies.”. Presumably the lack of oxygen at the altitude of Chamonix…. Given an “initial guess” at the prior, g, a conjugate prior (in dimension one with an invertible cdf), a family of priors is created in what first looks like a form of non-parametric exponential tilting of g. But a closer look [at (2.1)] exposes the “family” as the tautological π(θ)=g(θ)x π(θ)/g(θ). The ratio is expanded into a Legendre polynomial series. Which use in Bayesian statistics dates a wee bit further back than indicated in the paper (see, e.g., Friedman, 1985; Diaconis, 1986). With the side issue that the resulting approximation does not integrate to one. Another side issue is that the coefficients of the Legendre truncated series are approximated by simulations from the prior [Step 3 of the Type II algorithm], rarely an efficient approach to the posterior.

inferential models: reasoning with uncertainty [book review]

“the field of statistics (…) is still surprisingly underdeveloped (…) the subject lacks a solid theory for reasoning with uncertainty [and] there has been very little progress on the foundations of statistical inference” (p.xvi)

A book that starts with such massive assertions is certainly hoping to attract some degree of attention from the field and likely to induce strong reactions to this dismissal of the not inconsiderable amount of research dedicated so far to statistical inference and in particular to its foundations. Or even attarcting flak for not accounting (in this introduction) for the past work of major statisticians, like Fisher, Kiefer, Lindley, Cox, Berger, Efron, Fraser and many many others…. Judging from the references and the tone of this 254 pages book, it seems like the two authors, Ryan Martin and Chuanhai Liu, truly aim at single-handedly resetting the foundations of statistics to their own tune, which sounds like a new kind of fiducial inference augmented with calibrated belief functions. Be warned that five chapters of this book are built on as many papers written by the authors in the past three years. Which makes me question, if I may, the relevance of publishing a book on a brand-new approach to statistics without further backup from a wider community.

“…it is possible to calibrate our belief probabilities for a common interpretation by intelligent minds.” (p.14)

Chapter 1 contains a description of the new perspective in Section 1.4.2, which I find useful to detail here. When given an observation x from a Normal N(θ,1) model, the authors rewrite X as θ+Z, with Z~N(0,1), as in fiducial inference, and then want to find a “meaningful prediction of Z independently of X”. This seems difficult to accept given that, once X=x is observed, Z=X-θ⁰, θ⁰ being the true value of θ, which belies the independence assumption. The next step is to replace Z~N(0,1) by a random set S(Z) containing Z and to define a belief function bel() on the parameter space Θ by

bel(A|X) = P(X-S(Z)⊆A)

which induces a pseudo-measure on Θ derived from the distribution of an independent Z, since X is already observed. When Z~N(0,1), this distribution does not depend on θ⁰ the true value of θ… The next step is to choose the belief function towards a proper frequentist coverage, in the approximate sense that the probability that bel(A|X) be more than 1-α is less than α when the [arbitrary] parameter θ is not in A. And conversely. This property (satisfied when bel(A|X) is uniform) is called validity or exact inference by the authors: in my opinion, restricted frequentist calibration would certainly sound more adequate.

“When there is no prior information available, [the philosophical justifications for Bayesian analysis] are less than fully convincing.” (p.30)

“Is it logical that an improper “ignorance” prior turns into a proper “non-ignorance” prior when combined with some incomplete information on the whereabouts of θ?” (p.44)