Archive for Royal Statistical Society

a Bayesian criterion for singular models [discussion]

Posted in Books, Statistics, University life with tags , , , , , , , , , , , , , , , , on October 10, 2016 by xi'an

London Docks 12/02/09[Here is the discussion Judith Rousseau and I wrote about the paper by Mathias Drton and Martyn Plummer, a Bayesian criterion for singular models, which was discussed last week at the Royal Statistical Society. There is still time to send a written discussion! Note: This post was written using the latex2wp converter.]

It is a well-known fact that the BIC approximation of the marginal likelihood in a given irregular model {\mathcal M_k} fails or may fail. The BIC approximation has the form

\displaystyle BIC_k = \log p(\mathbf Y_n| \hat \pi_k, \mathcal M_k) - d_k \log n /2

where {d_k } corresponds on the number of parameters to be estimated in model {\mathcal M_k}. In irregular models the dimension {d_k} typically does not provide a good measure of complexity for model {\mathcal M_k}, at least in the sense that it does not lead to an approximation of

\displaystyle \log m(\mathbf Y_n |\mathcal M_k) = \log \left( \int_{\mathcal M_k} p(\mathbf Y_n| \pi_k, \mathcal M_k) dP(\pi_k|k )\right) \,.

A way to understand the behaviour of {\log m(\mathbf Y_n |\mathcal M_k) } is through the effective dimension

\displaystyle \tilde d_k = -\lim_n \frac{ \log P( \{ KL(p(\mathbf Y_n| \pi_0, \mathcal M_k) , p(\mathbf Y_n| \pi_k, \mathcal M_k) ) \leq 1/n | k ) }{ \log n}

when it exists, see for instance the discussions in Chambaz and Rousseau (2008) and Rousseau (2007). Watanabe (2009} provided a more precise formula, which is the starting point of the approach of Drton and Plummer:

\displaystyle \log m(\mathbf Y_n |\mathcal M_k) = \log p(\mathbf Y_n| \hat \pi_k, \mathcal M_k) - \lambda_k(\pi_0) \log n + [m_k(\pi_0) - 1] \log \log n + O_p(1)

where {\pi_0} is the true parameter. The authors propose a clever algorithm to approximate of the marginal likelihood. Given the popularity of the BIC criterion for model choice, obtaining a relevant penalized likelihood when the models are singular is an important issue and we congratulate the authors for it. Indeed a major advantage of the BIC formula is that it is an off-the-shelf crierion which is implemented in many softwares, thus can be used easily by non statisticians. In the context of singular models, a more refined approach needs to be considered and although the algorithm proposed by the authors remains quite simple, it requires that the functions { \lambda_k(\pi)} and {m_k(\pi)} need be known in advance, which so far limitates the number of problems that can be thus processed. In this regard their equation (3.2) is both puzzling and attractive. Attractive because it invokes nonparametric principles to estimate the underlying distribution; puzzling because why should we engage into deriving an approximation like (3.1) and call for Bayesian principles when (3.1) is at best an approximation. In this case why not just use a true marginal likelihood?

1. Why do we want to use a BIC type formula?

The BIC formula can be viewed from a purely frequentist perspective, as an example of penalised likelihood. The difficulty then stands into choosing the penalty and a common view on these approaches is to choose the smallest possible penalty that still leads to consistency of the model choice procedure, since it then enjoys better separation rates. In this case a {\log \log n} penalty is sufficient, as proved in Gassiat et al. (2013). Now whether or not this is a desirable property is entirely debatable, and one might advocate that for a given sample size, if the data fits the smallest model (almost) equally well, then this model should be chosen. But unless one is specifying what equally well means, it does not add much to the debate. This also explains the popularity of the BIC formula (in regular models), since it approximates the marginal likelihood and thus benefits from the Bayesian justification of the measure of fit of a model for a given data set, often qualified of being a Bayesian Ockham’s razor. But then why should we not compute instead the marginal likelihood? Typical answers to this question that are in favour of BIC-type formula include: (1) BIC is supposingly easier to compute and (2) BIC does not call for a specification of the prior on the parameters within each model. Given that the latter is a difficult task and that the prior can be highly influential in non-regular models, this may sound like a good argument. However, it is only apparently so, since the only justification of BIC is purely asymptotic, namely, in such a regime the difficulties linked to the choice of the prior disappear. This is even more the case for the sBIC criterion, since it is only valid if the parameter space is compact. Then the impact of the prior becomes less of an issue as non informative priors can typically be used. With all due respect, the solution proposed by the authors, namely to use the posterior mean or the posterior mode to allow for non compact parameter spaces, does not seem to make sense in this regard since they depend on the prior. The same comments apply to the author’s discussion on Prior’s matter for sBIC. Indeed variations of the sBIC could be obtained by penalizing for bigger models via the prior on the weights, for instance as in Mengersen and Rousseau (2011) or by, considering repulsive priors as in Petralia et al. (20120, but then it becomes more meaningful to (again) directly compute the marginal likelihood. Remains (as an argument in its favour) the relative computational ease of use of sBIC, when compared with the marginal likelihood. This simplification is however achieved at the expense of requiring a deeper knowledge on the behaviour of the models and it therefore looses the off-the-shelf appeal of the BIC formula and the range of applications of the method, at least so far. Although the dependence of the approximation of {\log m(\mathbf Y_n |\mathcal M_k)} on {\mathcal M_j }, $latex {j \leq k} is strange, this does not seem crucial, since marginal likelihoods in themselves bring little information and they are only meaningful when compared to other marginal likelihoods. It becomes much more of an issue in the context of a large number of models.

2. Should we care so much about penalized or marginal likelihoods ?

Marginal or penalized likelihoods are exploratory tools in a statistical analysis, as one is trying to define a reasonable model to fit the data. An unpleasant feature of these tools is that they provide numbers which in themselves do not have much meaning and can only be used in comparison with others and without any notion of uncertainty attached to them. A somewhat richer approach of exploratory analysis is to interrogate the posterior distributions by either varying the priors or by varying the loss functions. The former has been proposed in van Havre et l. (2016) in mixture models using the prior tempering algorithm. The latter has been used for instance by Yau and Holmes (2013) for segmentation based on Hidden Markov models. Introducing a decision-analytic perspective in the construction of information criteria sounds to us like a reasonable requirement, especially when accounting for the current surge in studies of such aspects.

[Posted as arXiv:1610.02503]

[Royal] Series B’log

Posted in Books, Statistics, University life, Wines with tags , , , , , , on September 12, 2016 by xi'an

[Thanks to Ingmar for suggesting the additional Royal!]

RSS wineLast week, I got an email from Piotr Fryzlewicz on behalf of the Publication Committee of the Royal Statistical Society enquiring about my interest in becoming a blog associate editor for Series B! Although it does not come exactly as a surprise, as I had previously heard about this interest in creating a dedicated blog, this is great news as I think a lively blog can only enhance the visibility and impact of papers published in Series B and hence increase the influence of Series B. Being quite excited by this on-line and interactive extension to the journal, I have accepted the proposal and we are now working on designing the new blog (Series B’log!) to get it on track as quickly as possible.

Suggestions towards this experiment are most welcome! I am thinking of involving authors to write blog summaries of their paper, AEs and reviewers to voice their expert opinions about the paper, anonymously or not, and of course anyone interested in commenting the paper. The idea is to turn (almost) all papers into on-line Read Papers, with hopefully the backup of authors through their interactions with the commentators. I certainly do not intend to launch discussions on each and every paper, betting on the AEs or referees to share their impressions. And if a paper ends up being un-discussed, this may prove enough of an incentive for some. (Someone asked me if we intended to discuss rejected papers as well. This is an interesting concept, but not to be considered at the moment!)

a mistake in a 1990 paper

Posted in Kids, Statistics, University life with tags , , , , , , , , on August 7, 2016 by xi'an

As we were working on the Handbook of mixture analysis with Sylvia Früwirth-Schnatter and Gilles Celeux today, near Saint-Germain des Près, I realised that there was a mistake in our 1990 mixture paper with Jean Diebolt [published in 1994], in that when we are proposing to use improper “Jeffreys” priors under the restriction that no component of the Gaussian mixture is “empty”, meaning that there are at least two observations generated from each component, the likelihood needs to be renormalised to be a density for the sample. This normalisation constant only depends on the weights of the mixture, which means that, when simulating from the full conditional distribution of the weights, there should be an extra-acceptance step to account for this correction. Of course, the term is essentially equal to one for a large enough sample but this remains a mistake nonetheless! It is funny that it remained undetected for so long in my most cited paper. Checking on Larry’s 1999 paper exploring the idea of excluding terms from the likelihood to allow for improper priors, I did not spot him using a correction either.

the end of Series B!

Posted in Books, pictures, Statistics, University life with tags , , , , on May 25, 2016 by xi'an

I received this news from the RSS today that all the RSS journals are turning 100% electronic. No paper version any longer! I deeply regret this move on which, as an RSS member, I would have appreciated to be consulted as I find much easier to browse through the current issue when it arrives in my mailbox, rather than being t best reminded by an email that I will most likely ignore and erase. And as I consider the production of the journals the prime goal of the Royal Statistical Society. And as I read that only 25% of the members had opted so far for the electronic format, which does not sound to me like a majority. In addition, moving to electronic-only journals does not bring the perks one would expect from electronic journals:

  • no bonuses like supplementary material, code, open or edited comments
  • no reduction in the subscription rate of the journals and penalty fees if one still wants a paper version, which amounts to a massive increase in the subscription price
  • no disengagement from the commercial publisher, whose role become even less relevant
  • no access to the issues of the years one has paid for, once one stops subscribing.

“The benefits of electronic publishing include: faster publishing speeds; increased content; instant access from a range of electronic devices; additional functionality; and of course, environmental sustainability.”

The move is sold with typical marketing noise. But I do not buy it: publishing speeds will remain the same as driven by the reviewing part, I do not see where the contents are increased, and I cannot seriously read a journal article from my phone, so this range of electronic devices remains a gadget. Not happy!

the last digit of e

Posted in Kids, Mountains, pictures, Statistics, Travel, University life with tags , , , , , , , on March 3, 2016 by xi'an

Éric Marchand from Sherbrooke, Québec [historical birthplace of MCMC, since Adrian Smith gave his first talk on his Gibbs sampler there, in June 1989], noticed my recent posts about the approximation of e by Monte Carlo methods and sent me a paper he wrote in The Mathematical Gazette of November 1995 [full MCMC era!] about original proofs on the expectation of some stopping rules being e, like the length of increasing runs. And Gnedenko’s uniform summation until exceeding one. Amazing that this simple problem generated so much investigation!!!


Posted in Books, Statistics with tags , , , , , , , , on December 1, 2015 by xi'an

WariseWhile in Warwick this week, I borrowed a recent issue (Oct. 08, 2015) of Nature from Tom Nichols and read it over diners in a maths house. Its featured topic was reproducibility, with a long initial (or introductory) article about “Fooling ourselves”, starting with an illustration from Andrew himself who had gotten a sign wrong in one of those election studies that are the basis of Red State, Blue State. While this article is not bringing radically new perspectives on the topic, there is nothing shocking about it and it even goes on mentioning Peter Green and his Royal Statistical Society President’s tribune about the Sally Clark case and Eric-Jan Wagenmakers with a collaboration with competing teams that sounded like “putting one’s head on a guillotine”. Which relates to a following “comment” on crowdsourcing research or data analysis.

I however got most interested by another comment by MacCoun and Perlmutter, where they advocate a systematic blinding of data to avoid conscious or unconscious biases. While I deem the idea quite interesting and connected with anonymisation techniques in data privacy, I find the presentation rather naïve in its goals (from a statistical perspective). Indeed, if we consider data produced by a scientific experiment towards the validation or invalidation of a scientific hypothesis, it usually stands on its own, with no other experiment of a similar kind to refer to. Add too much noise and only noise remains. Add too little and the original data remains visible. This means it is quite difficult to calibrate the blinding mechanisms in order for the blinded data to remain realistic enough to be analysed. Or to be different enough from the original data for different conclusions to be drawn. The authors suggest blinding being done by a software, by adding noise, bias, label switching, &tc. But I do not think this blinding can be done blindly, i.e., without a clear idea of what the possible models are, so that the perturbed datasets created out of the original data favour more one of the models under comparison. And are realistic for at least one of those models. Thus, some preliminary analysis of the original or of some pseudo-data from each of the proposed models is somewhat unavoidable to calibrate the blinding machinery towards realistic values. If designing a new model is part of the inferential goals, this may prove impossible… Again, I think having several analyses run in parallel with several perturbed datasets quite a good idea to detect the impact of some prior assumptions. But this requires statistically savvy programmers. And possibly informative prior distributions.

World statistics day

Posted in pictures, Statistics, Travel, University life with tags , , , , , , , , , on October 20, 2015 by xi'an

Today is October 20, World Statistics Day as launched by the UN. And supported by local and international societies. In connection with that day, among many events, the RSS will be hosting a reception, China will hold a seminar in… Xi’an, how appropriate!, my friend Kerrie Mengersen will give a talk at the Queensland University of Technology (QUT) on The power and promise of immersive virtual environments. (Bringing her pet crocodile to the talk, hopefully!)

Leuven9And I will also give a talk in Louvain-la-Neuve, Belgium, on Le délicat dilemme des tests d’hypothèse et de leur résolution bayésienne. At ISBA, which stands for Institute of Statistics, Biostatistics and Actuarial Sciences and not for the Bayesian society!. within UCL, which stands for Université Catholique de Louvain and not for University College London! (And which is not to be confused with the Katholieke Universiteit Leuven, in Leuven, where I was last year for MCqMC. About 25 kilometers away.) In case this is not confusing enough, here are my slides (in English, while the talk will be in French):