mea culpa!

An entry about our Bayesian Essentials book on X validated alerted me to a typo in the derivation of the Gaussian posterior..! When deriving the posterior (which was left as an exercise in the Bayesian Core), I just forgot the term expressing the divergence between the prior mean and the sample mean. Mea culpa!!!

priors without likelihoods are like sloths without…

“The idea of building priors that generate reasonable data may seem like an unusual idea…”

Andrew, Dan, and Michael arXived a opinion piece last week entitled “The prior can generally only be understood in the context of the likelihood”. Which connects to the earlier Read Paper of Gelman and Hennig I discussed last year. I cannot state strong disagreement with the positions taken in this piece, actually, in that I do not think prior distributions ever occur as a given but are rather chosen as a reference measure to probabilise the parameter space and eventually prioritise regions over others. If anything I find myself even further on the prior agnosticism gradation.  (Of course, this lack of disagreement applies to the likelihood understood as a function of both the data and the parameter, rather than of the parameter only, conditional on the data. Priors cannot be depending on the data without incurring disastrous consequences!)

“…it contradicts the conceptual principle that the prior distribution should convey only information that is available before the data have been collected.”

The first example is somewhat disappointing in that it revolves as so many Bayesian textbooks (since Laplace!) around the [sex ratio] Binomial probability parameter and concludes at the strong or long-lasting impact of the Uniform prior. I do not see much of a contradiction between the use of a Uniform prior and the collection of prior information, if only because there is not standardised way to transfer prior information into prior construction. And more fundamentally because a parameter rarely makes sense by itself, alone, without a model that relates it to potential data. As for instance in a regression model. More, following my epiphany of last semester, about the relativity of the prior, I see no damage in the prior being relevant, as I only attach a relative meaning to statements based on the posterior. Rather than trying to limit the impact of a prior, we should rather build assessment tools to measure this impact, for instance by prior predictive simulations. And this is where I come to quite agree with the authors.

“…non-identifiabilities, and near nonidentifiabilites, of complex models can lead to unexpected amounts of weight being given to certain aspects of the prior.”

Another rather straightforward remark is that non-identifiable models see the impact of a prior remain as the sample size grows. And I still see no issue with this fact in a relative approach. When the authors mention (p.7) that purely mathematical priors perform more poorly than weakly informative priors it is hard to see what they mean by this “performance”.

“…judge a prior by examining the data generating processes it favors and disfavors.”

Besides those points, I completely agree with them about the fundamental relevance of the prior as a generative process, only when the likelihood becomes available. And simulatable. (This point is found in many references, including our response to the American Statistician paper Hidden dangers of specifying noninformative priors, with Kaniav Kamary. With the same illustration on a logistic regression.) I also agree to their criticism of the marginal likelihood and Bayes factors as being so strongly impacted by the choice of a prior, if treated as absolute quantities. I also if more reluctantly and somewhat heretically see a point in using the posterior predictive for assessing whether a prior is relevant for the data at hand. At least at a conceptual level. I am however less certain about how to handle improper priors based on their recommendations. In conclusion, it would be great to see one [or more] of the authors at O-Bayes 2017 in Austin as I am sure it would stem nice discussions there! (And by the way I have no prior idea on how to conclude the comparison in the title!)

better together?

Yesterday came out on arXiv a joint paper by Pierre Jacob, Lawrence Murray, Chris Holmes and myself, Better together? Statistical learning in models made of modules, paper that was conceived during the MCMski meeting in Chamonix, 2014! Indeed it is mostly due to Martyn Plummer‘s talk at this meeting about the cut issue that we started to work on this topic at the fringes of the [standard] Bayesian world. Fringes because a standard Bayesian approach to the problem would always lead to use the entire dataset and the entire model to infer about a parameter of interest. [Disclaimer: the use of the very slogan of the anti-secessionists during the Scottish Independence Referendum of 2014 in our title is by no means a measure of support of their position!] Comments and suggested applications most welcomed!

The setting of the paper is inspired by realistic situations where a model is made of several modules, connected within a graphical model that represents the statistical dependencies, each relating to a specific data modality. In a standard Bayesian analysis, given data, a conventional statistical update then allows for coherent uncertainty quantification and information propagation through and across the modules. However, misspecification of or even massive uncertainty about any module in the graph can contaminate the estimate and update of parameters of other modules, often in unpredictable ways. Particularly so when certain modules are trusted more than others. Hence the appearance of cut models, where practitioners  prefer skipping the full model and limit the information propagation between these modules, for example by restricting propagation to only one direction along the edges of the graph. (Which is sometimes represented as a diode on the edge.) The paper investigates in which situations and under which formalism such modular approaches can outperform the full model approach in misspecified settings. By developing the appropriate decision-theoretic framework. Meaning we can choose between [several] modular and full-model approaches.

Jeffreys priors for mixtures [or not]

Clara Grazian and I have just arXived [and submitted] a paper on the properties of Jeffreys priors for mixtures of distributions. (An earlier version had not been deemed of sufficient interest by Bayesian Analysis.) In this paper, we consider the formal Jeffreys prior for a mixture of Gaussian distributions and examine whether or not it leads to a proper posterior with a sufficient number of observations.  In general, it does not and hence cannot be used as a reference prior. While this is a negative result (and this is why Bayesian Analysis did not deem it of sufficient importance), I find it definitely relevant because it shows that the default reference prior [in the sense that the Jeffreys prior is the primary choice in nonparametric settings] does not operate in this wide class of distributions. What is surprising is that the use of a Jeffreys-like prior on a global location-scale parameter (as in our 1996 paper with Kerrie Mengersen or our recent work with Kaniav Kamary and Kate Lee) remains legit if proper priors are used on all the other parameters. (This may be yet another illustration of the tequilla-like toxicity of mixtures!)

Francisco Rubio and Mark Steel already exhibited this difficulty of the Jeffreys prior for mixtures of densities with disjoint supports [which reveals the mixture latent variable and hence turns the problem into something different]. Which relates to another point of interest in the paper, derived from a 1988 [Valencià Conference!] paper by José Bernardo and Javier Giròn, where they show the posterior associated with a Jeffreys prior on a mixture is proper when (a) only estimating the weights p and (b) using densities with disjoint supports. José and Javier use in this paper an astounding argument that I had not seen before and which took me a while to ingest and accept. Namely, the Jeffreys prior on a observed model with latent variables is bounded from above by the Jeffreys prior on the corresponding completed model. Hence if the later leads to a proper posterior for the observed data, so does the former. Very smooth, indeed!!!

Actually, we still support the use of the Jeffreys prior but only for the mixture mixtures, because it has the property supported by Judith and Kerrie of a conservative prior about the number of components. Obviously, we cannot advocate its use over all the parameters of the mixture since it then leads to an improper posterior.

what makes variables randoms [book review]

When the goal of a book is to make measure theoretic probability available to applied researchers for conducting their research, I cannot but applaud! Peter Veazie’s goal of writing “a brief text that provides a basic conceptual introduction to measure theory” (p.4) is hence most commendable. Before reading What makes variables random, I was uncertain how this could be achieved with a limited calculus background, given the difficulties met by our third year maths students. After reading the book, I am even less certain this is feasible!

“…it is the data generating process that makes the variables random and not the data.”

Chapter 2 is about basic notions of set theory. Chapter 3 defines measurable sets and measurable functions and integrals against a given measure μ as

which I find particularly unnatural compared with the definition through simple functions (esp. because it does not tell how to handle 0x∞). The ensuing discussion shows the limitation of the exercise in that the definition is only explained for finite sets (since the notion of a partition achieving the supremum on page 29 is otherwise meaningless). A generic problem with the book, in that most examples in the probability section relate to discrete settings (see the discussion of the power set p.66). I also did not see a justification as to why measurable functions enjoy well-defined integrals in the above sense. All in all, to see less than ten pages allocated to measure theory per se is rather staggering! For instance,

does not appear to be defined at all.

“…the mathematical probability theory underlying our analyses is just mathematics…”

Chapter 4 moves to probability measures. It distinguishes between objective (or frequentist) and subjective measures, which is of course open to diverse interpretations. And the definition of a conditional measure is the traditional one, conditional on a set rather than on a σ-algebra. Surprisingly as this is in my opinion one major reason for using measures in probability theory. And avoids unpleasant issues such as Bertrand’s paradox. While random variables are defined in the standard sense of real valued measurable functions, I did not see a definition of a continuous random variables or of the Lebesgue measure. And there are only a few lines (p.48) about the notion of expectation, which is so central to measure-theoretic probability as to provide a way of entry into measure theory! Progressing further, the σ-algebra induced by a random variable is defined as a partition (p.52), a particularly obscure notion for continuous rv’s. When the conditional density of one random variable given the realisation of another is finally introduced (p.63), as an expectation reconciling with the set-wise definition of conditional probabilities, it is in a fairly convoluted way that I fear will scare newcomers out of their wit. Since it relies on a sequence of nested sets with positive measure, implying an underlying topology and the like, which somewhat shows the impossibility of the overall task…

“In the Bayesian analysis, the likelihood provides meaning to the posterior.”

Statistics is hurriedly introduced in a short section at the end of Chapter 4, assuming the notion of likelihood is already known by the readers. But nitpicking (p.65) at the representation of the terms in the log-likelihood as depending on an unspecified parameter value θ [not to be confused with the data-generating value of θ, which does not appear clearly in this section]. Section that manages to include arcane remarks distinguishing maximum likelihood estimation from Bayesian analysis, all this within a page! (Nowhere is the Bayesian perspective clearly defined.)

“We should no more perform an analysis clustered by state than we would cluster by age, income, or other random variable.”

The last part of the book is about probabilistic models, drawing a distinction between data generating process models and data models (p.89), by which the author means the hypothesised probabilistic model versus the empirical or bootstrap distribution. An interesting way to relate to the main thread, except that the convergence of the data distribution to the data generating process model cannot be established at this level. And hence that the very nature of bootstrap may be lost on the reader. A second and final chapter covers some common or vexing problems and the author’s approach to them. Revolving around standard errors, fixed and random effects. The distinction between standard deviation (“a mathematical property of a probability distribution”) and standard error (“representation of variation due to a data generating process”) that is followed for several pages seems to boil down to a possible (and likely) model mis-specification. The chapter also contains an extensive discussion of notations, like indexes (or indicators), which seems a strange focus esp. at this location in the book. Over 15 pages! (Furthermore, I find quite confusing that a set of indices is denoted there by the double barred I, usually employed for the indicator function.)

“…the reader will probably observe the conspicuous absence of a time-honoured topic in calculus courses, the “Riemann integral”… Only the stubborn conservatism of academic tradition could freeze it into a regular part of the curriculum, long after it had outlived its historical importance.” Jean Dieudonné, Foundations of Modern Analysis

In conclusion, I do not see the point of this book, from its insistence on measure theory that never concretises for lack of mathematical material to an absence of convincing examples as to why this is useful for the applied researcher, to the intended audience which is expected to already quite a lot about probability and statistics, to a final meandering around linear models that seems at odds with the remainder of What makes variables random, without providing an answer to this question. Or to the more relevant one of why Lebesgue integration is preferable to Riemann integration. (Not that there does not exist convincing replies to this question!)

Fourth Bayesian, Fiducial, and Frequentist Conference

Next May 1-3, I will attend the 4th Bayesian, Fiducial and Frequentist Conference at Harvard University (hopefully not under snow at that time of year), which is a meeting between philosophers and statisticians about foundational thinking in statistics and inference under uncertainty. This should be fun! (Registration is now open.)

X-Outline of a Theory of Statistical Estimation

While visiting Warwick last week, Jean-Michel Marin pointed out and forwarded me this remarkable paper of Jerzy Neyman, published in 1937, and presented to the Royal Society by Harold Jeffreys.

“Leaving apart on one side the practical difficulty of achieving randomness and the meaning of this word when applied to actual experiments…”

“It may be useful to point out that although we are frequently witnessing controversies in which authors try to defend one or another system of the theory of probability as the only legitimate, I am of the opinion that several such theories may be and actually are legitimate, in spite of their occasionally contradicting one another. Each of these theories is based on some system of postulates, and so long as the postulates forming one particular system do not contradict each other and are sufficient to construct a theory, this is as legitimate as any other. “

This paper is fairly long in part because Neyman starts by setting Kolmogorov’s axioms of probability. This is of historical interest but also needed for Neyman to oppose his notion of probability to Jeffreys’ (which is the same from a formal perspective, I believe!). He actually spends a fair chunk on explaining why constants cannot have anything but trivial probability measures. Getting ready to state that an a priori distribution has no meaning (p.343) and that in the rare cases it does it is mostly unknown. While reading the paper, I thought that the distinction was more in terms of frequentist or conditional properties of the estimators, Neyman’s arguments paving the way to his definition of a confidence interval. Assuming repeatability of the experiment under the same conditions and therefore same parameter value (p.344).

“The advantage of the unbiassed [sic] estimates and the justification of their use lies in the fact that in cases frequently met the probability of their differing very much from the estimated parameters is small.”

“…the maximum likelihood estimates appear to be what could be called the best “almost unbiassed [sic]” estimates.”

It is also quite interesting to read that the principle for insisting on unbiasedness is one of producing small errors, because this is not that often the case, as shown by the complete class theorems of Wald (ten years later). And that maximum likelihood is somewhat relegated to a secondary rank, almost unbiased being understood as consistent. A most amusing part of the paper is when Neyman inverts the credible set into a confidence set, that is, turning what is random in a constant and vice-versa. With a justification that the credible interval has zero or one coverage, while the confidence interval has a long-run validity of returning the correct rate of success. What is equally amusing is that the boundaries of a credible interval turn into functions of the sample, hence could be evaluated on a frequentist basis, as done later by Dennis Lindley and others like Welch and Peers, but that Neyman fails to see this and turn the bounds into hard values. For a given sample.

“This, however, is not always the case, and in general there are two or more systems of confidence intervals possible corresponding to the same confidence coefficient α, such that for certain sample points, E’, the intervals in one system are shorter than those in the other, while for some other sample points, E”, the reverse is true.”

The resulting construction of a confidence interval is then awfully convoluted when compared with the derivation of an HPD region, going through regions of acceptance that are the dual of a confidence interval (in the sampling space), while apparently [from my hasty read] missing a rule to order them. And rejecting the notion of a confidence interval being possibly empty, which, while being of practical interest, clashes with its frequentist backup.