Archive for Bayes factor

are there a frequentist and a Bayesian likelihoods?

Posted in Statistics with tags , , , , , , , , , , on June 7, 2018 by xi'an

A question that came up on X validated and led me to spot rather poor entries in Wikipedia about both the likelihood function and Bayes’ Theorem. Where unnecessary and confusing distinctions are made between the frequentist and Bayesian versions of these notions. I have already discussed the later (Bayes’ theorem) a fair amount here. The discussion about the likelihood is quite bemusing, in that the likelihood function is the … function of the parameter equal to the density indexed by this parameter at the observed value.

“What we can find from a sample is the likelihood of any particular value of r, if we define the likelihood as a quantity proportional to the probability that, from a population having the particular value of r, a sample having the observed value of r, should be obtained.” R.A. Fisher, On the “probable error’’ of a coefficient of correlation deduced from a small sample. Metron 1, 1921, p.24

By mentioning an informal side to likelihood (rather than to likelihood function), and then stating that the likelihood is not a probability in the frequentist version but a probability in the Bayesian version, the W page makes a complete and unnecessary mess. Whoever is ready to rewrite this introduction is more than welcome! (Which reminded me of an earlier question also on X validated asking why a common reference measure was needed to define a likelihood function.)

This also led me to read a recent paper by Alexander Etz, whom I met at E.J. Wagenmakers‘ lab in Amsterdam a few years ago. Following Fisher, as Jeffreys complained about

“..likelihood, a convenient term introduced by Professor R.A. Fisher, though in his usage it is sometimes multiplied by a constant factor. This is the probability of the observations given the original information and the hypothesis under discussion.” H. Jeffreys, Theory of Probability, 1939, p.28

Alexander defines the likelihood up to a constant, which causes extra-confusion, for free!, as there is no foundational reason to introduce this degree of freedom rather than imposing an exact equality with the density of the data (albeit with an arbitrary choice of dominating measure, never neglect the dominating measure!). The paper also repeats the message that the likelihood is not a probability (density, missing in the paper). And provides intuitions about maximum likelihood, likelihood ratio and Wald tests. But does not venture into a separate definition of the likelihood, being satisfied with the fundamental notion to be plugged into the magical formula

posteriorprior×likelihood

Practicals of Uncertainty [book review]

Posted in Books, Statistics, University life with tags , , , , , , , on December 22, 2017 by xi'an

On my way to the O’Bayes 2017 conference in Austin, I [paradoxically!] went through Jay Kadane’s Pragmatics of Uncertainty, which had been published earlier this year by CRC Press. The book is to be seen as a practical illustration of the Principles of Uncertainty Jay wrote in 2011 (and I reviewed for CHANCE). The avowed purpose is to allow the reader to check through Jay’s applied work whether or not he had “made good” on setting out clearly the motivations for his subjective Bayesian modelling. (While I presume the use of the same P of U in both books is mostly a coincidence, I started wondering how a third P of U volume could be called. Perils of Uncertainty? Peddlers of Uncertainty? The game is afoot!)

The structure of the book is a collection of fifteen case studies undertaken by Jay over the past 30 years, covering paleontology, survey sampling, legal expertises, physics, climate, and even medieval Norwegian history. Each chapter starts with a short introduction that often explains how he came by the problem (most often as an interesting PhD student consulting project at CMU), what were the difficulties in the analysis, and what became of his co-authors. As noted by the author, the main bulk of each chapter is the reprint (in a unified style) of the paper and most of these papers are actually and freely available on-line. The chapter always concludes with an epilogue (or post-mortem) that re-considers (very briefly) what had been done and what could have been done and whether or not the Bayesian perspective was useful for the problem (unsurprisingly so for the majority of the chapters!). There are also reading suggestions in the other P of U and a few exercises.

“The purpose of the book is philosophical, to address, with specific examples, the question of whether Bayesian statistics is ready for prime time. Can it be used in a variety of applied settings to address real applied problems?”

The book thus comes as a logical complement of the Principles, to demonstrate how Jay himself did apply his Bayesian principles to specific cases and how one can set the construction of a prior, of a loss function or of a statistical model in identifiable parts that can then be criticised or reanalysed. I find browsing through this series of fourteen different problems fascinating and exhilarating, while I admire the dedication of Jay to every case he presents in the book. I also feel that this comes as a perfect complement to the earlier P of U, in that it makes refering to a complete application of a given principle most straightforward, the problem being entirely described, analysed, and in most cases solved within a given chapter. A few chapters have discussions, being published in the Valencia meeting proceedings or another journal with discussions.

While all papers have been reset in the book style, I wish the graphs had been edited as well as they do not always look pretty. Although this would have implied a massive effort, it would have also been great had each chapter and problem been re-analysed or at least discussed by another fellow (?!) Bayesian in order to illustrate the impact of individual modelling sensibilities. This may however be a future project for a graduate class. Assuming all datasets are available, which is unclear from the text.

“We think however that Bayes factors are overemphasized. In the very special case in which there are only two possible “states of the world”, Bayes factors are sufficient. However in the typical case in which there are many possible states of the world, Bayes factors are sufficient only when the decision-maker’s loss has only two values.” (p. 278)

The above is in Jay’s reply to a comment from John Skilling regretting the absence of marginal likelihoods in the chapter. Reply to which I completely subscribe.

[Usual warning: this review should find its way into CHANCE book reviews at some point, with a fairly similar content.]

the Hyvärinen score is back

Posted in pictures, Statistics, Travel with tags , , , , , , , , , , , , , on November 21, 2017 by xi'an

Stéphane Shao, Pierre Jacob and co-authors from Harvard have just posted on arXiv a new paper on Bayesian model comparison using the Hyvärinen score

\mathcal{H}(y, p) = 2\Delta_y \log p(y) + ||\nabla_y \log p(y)||^2

which thus uses the Laplacian as a natural and normalisation-free penalisation for the score test. (Score that I first met in Padova, a few weeks before moving from X to IX.) Which brings a decision-theoretic alternative to the Bayes factor and which delivers a coherent answer when using improper priors. Thus a very appealing proposal in my (biased) opinion! The paper is mostly computational in that it proposes SMC and SMC² solutions to handle the estimation of the Hyvärinen score for models with tractable likelihoods and tractable completed likelihoods, respectively. (Reminding me that Pierre worked on SMC² algorithms quite early during his Ph.D. thesis.)

A most interesting remark in the paper is to recall that the Hyvärinen score associated with a generic model on a series must be the prequential (predictive) version

\mathcal{H}_T (M) = \sum_{t=1}^T \mathcal{H}(y_t; p_M(dy_t|y_{1:(t-1)}))

rather than the version on the joint marginal density of the whole series. (Followed by a remark within the remark that the logarithm scoring rule does not make for this distinction. And I had to write down the cascading representation

\log p(y_{1:T})=\sum_{t=1}^T \log p(y_t|y_{1:t-1})

to convince myself that this unnatural decomposition, where the posterior on θ varies on each terms, is true!) For consistency reasons.

This prequential decomposition is however a plus in terms of computation when resorting to sequential Monte Carlo. Since each time step produces an evaluation of the associated marginal. In the case of state space models, another decomposition of the authors, based on measurement densities and partial conditional expectations of the latent states allows for another (SMC²) approximation. The paper also establishes that for non-nested models, the Hyvärinen score as a model selection tool asymptotically selects the closest model to the data generating process. For the divergence induced by the score. Even for state-space models, under some technical assumptions.  From this asymptotic perspective, the paper exhibits an example where the Bayes factor and the Hyvärinen factor disagree, even asymptotically in the number of observations, about which mis-specified model to select. And last but not least the authors propose and assess a discrete alternative relying on finite differences instead of derivatives. Which remains a proper scoring rule.

I am quite excited by this work (call me biased!) and I hope it can induce following works as a viable alternative to Bayes factors, if only for being more robust to the [unspecified] impact of the prior tails. As in the above picture where some realisations of the SMC² output and of the sequential decision process see the wrong model being almost acceptable for quite a long while…

WBIC, practically

Posted in Statistics with tags , , , , , , , , , on October 20, 2017 by xi'an

“Thus far, WBIC has received no more than a cursory mention by Gelman et al. (2013)”

I had missed this 2015  paper by Nial Friel and co-authors on a practical investigation of Watanabe’s WBIC. Where WBIC stands for widely applicable Bayesian information criterion. The thermodynamic integration approach explored by Nial and some co-authors for the approximation of the evidence, thermodynamic integration that produces the log-evidence as an integral between temperatures t=0 and t=1 of a powered evidence, is eminently suited for WBIC, as the widely applicable Bayesian information criterion is associated with the specific temperature t⁰ that makes the power posterior equidistant, Kullback-Leibler-wise, from the prior and posterior distributions. And the expectation of the log-likelihood under this very power posterior equal to the (genuine) evidence. In fact, WBIC is often associated with the sub-optimal temperature 1/log(n), where n is the (effective?) sample size. (By comparison, if my minimalist description is unclear!, thermodynamic integration requires a whole range of temperatures and associated MCMC runs.) In an ideal Gaussian setting, WBIC improves considerably over thermodynamic integration, the larger the sample the better. In more realistic settings, though, including a simple regression and a logistic [Pima Indians!] model comparison, thermodynamic integration may do better for a given computational cost although the paper is unclear about these costs. The paper also runs a comparison with harmonic mean and nested sampling approximations. Since the integral of interest involves a power of the likelihood, I wonder if a safe version of the harmonic mean resolution can be derived from simulations of the genuine posterior. Provided the exact temperature t⁰ is known…

how many academics does it take to change… a p-value threshold?

Posted in Books, pictures, Running, Statistics, Travel with tags , , , , , , , , on August 22, 2017 by xi'an

“…a critical mass of researchers now endorse this change.”

The answer to the lightpulp question seems to be 72: Andrew sent me a short paper recently PsyarXived and to appear in Nature Human Behaviour following on the .005 not .05 tune we criticised in PNAS a while ago. (Actually a very short paper once the names and affiliations of all authors are taken away.) With indeed 72 authors, many of them my Bayesian friends! I figure the mass signature is aimed at convincing users of p-values of a consensus among statisticians. Or a “critical mass” as stated in the note. On the next week, Nature had an entry on this proposal. (With a survey on whether the p-value threshold should change!)

The argument therein [and hence my reservations] is about the same as in Val Johnson’s original PNAS paper, namely that .005 should become the reference cutoff when using p-values for discovering new effects. The tone of the note is mostly Bayesian in that it defends the Bayes factor as a better alternative I would call the b-value. And produces graphs that relate p-values to some minimax Bayes factors. In the simplest possible case of testing for the nullity of a normal mean. Which I do not think is particularly convincing when considering more realistic settings with (many) nuisance parameters and possible latent variables where numerical answers diverge between p-values and [an infinity of] b-values. And of course the unsolved issue of scaling the Bayes factor. (This without embarking anew upon a full-fledged criticism of the Bayes factor.) As usual, I am also skeptical of mentions of power, since I never truly understood the point of power, which depends on the alternative model, increasingly so with the complexity of this alternative. As argued in our letter to PNAS, the central issue that this proposal fails to address is the urgency in abandoning the notion [indoctrinated in generations of students that a single quantity and a single bound are the answers to testing issues. Changing the bound sounds like suggesting to paint afresh a building on the verge of collapsing.

a response by Ly, Verhagen, and Wagenmakers

Posted in Statistics with tags , , , , , , , , on March 9, 2017 by xi'an

Following my demise [of the Bayes factor], Alexander Ly, Josine Verhagen, and Eric-Jan Wagenmakers wrote a very detailed response. Which I just saw the other day while in Banff. (If not in Schiphol, which would have been more appropriate!)

“In this rejoinder we argue that Robert’s (2016) alternative view on testing has more in common with Jeffreys’s Bayes factor than he suggests, as they share the same ‘‘shortcomings’’.”

Rather unsurprisingly (!), the authors agree with my position on the dangers to ignore decisional aspects when using the Bayes factor. A point of dissension is the resolution of the Jeffreys[-Lindley-Bartlett] paradox. One consequence derived by Alexander and co-authors is that priors should change between testing and estimating. Because the parameters have a different meaning under the null and under the alternative, a point I agree with in that these parameters are indexed by the model [index!]. But with which I disagree when arguing that the same parameter (e.g., a mean under model M¹) should have two priors when moving from testing to estimation. To state that the priors within the marginal likelihoods “are not designed to yield posteriors that are good for estimation” (p.45) amounts to wishful thinking. I also do not find a strong justification within the paper or the response about choosing an improper prior on the nuisance parameter, e.g. σ, with the same constant. Another a posteriori validation in my opinion. However, I agree with the conclusion that the Jeffreys paradox prohibits the use of an improper prior on the parameter being tested (or of the test itself). A second point made by the authors is that Jeffreys’ Bayes factor is information consistent, which is correct but does not solved my quandary with the lack of precise calibration of the object, namely that alternatives abound in a non-informative situation.

“…the work by Kamary et al. (2014) impressively introduces an alternative view on testing, an algorithmic resolution, and a theoretical justification.”

The second part of the comments is highly supportive of our mixture approach and I obviously appreciate very much this support! Especially if we ever manage to turn the paper into a discussion paper! The authors also draw a connection with Harold Jeffreys’ distinction between testing and estimation, based upon Laplace’s succession rule. Unbearably slow succession law. Which is well-taken if somewhat specious since this is a testing framework where a single observation can send the Bayes factor to zero or +∞. (I further enjoyed the connection of the Poisson-versus-Negative Binomial test with Jeffreys’ call for common parameters. And the supportive comments on our recent mixture reparameterisation paper with Kaniav Kamari and Kate Lee.) The other point that the Bayes factor is more sensitive to the choice of the prior (beware the tails!) can be viewed as a plus for mixture estimation, as acknowledged there. (The final paragraph about the faster convergence of the weight α is not strongly

relativity is the keyword

Posted in Books, Statistics, University life with tags , , , , , , , on February 1, 2017 by xi'an

St John's College, Oxford, Feb. 23, 2012As I was teaching my introduction to Bayesian Statistics this morning, ending up with the chapter on tests of hypotheses, I found reflecting [out loud] on the relative nature of posterior quantities. Just like when I introduced the role of priors in Bayesian analysis the day before, I stressed the relativity of quantities coming out of the BBB [Big Bayesian Black Box], namely that whatever happens as a Bayesian procedure is to be understood, scaled, and relativised against the prior equivalent, i.e., that the reference measure or gauge is the prior. This is sort of obvious, clearly, but bringing the argument forward from the start avoids all sorts of misunderstanding and disagreement, in that it excludes the claims of absolute and certainty that may come with the production of a posterior distribution. It also removes the endless debate about the determination of the prior, by making each prior a reference on its own. With an additional possibility of calibration by simulation under the assumed model. Or an alternative. Again nothing new there, but I got rather excited by this presentation choice, as it seems to clarify the path to Bayesian modelling and avoid misapprehensions.

Further, the curious case of the Bayes factor (or of the posterior probability) could possibly be resolved most satisfactorily in this framework, as the [dreaded] dependence on the model prior probabilities then becomes a matter of relativity! Those posterior probabilities depend directly and almost linearly on the prior probabilities, but they should not be interpreted in an absolute sense as the ultimate and unique probability of the hypothesis (which anyway does not mean anything in terms of the observed experiment). In other words, this posterior probability does not need to be scaled against a U(0,1) distribution. Or against the p-value if anyone wishes to do so. By the end of the lecture, I was even wondering [not so loudly] whether or not this perspective was allowing for a resolution of the Lindley-Jeffreys paradox, as the resulting number could be set relative to the choice of the [arbitrary] normalising constant. Continue reading