At Warwick U., CRiSM is sponsoring another workshop, this time on hypothesis testing next Sept. 15 and 16 (just before the Glencoe race!). Registration and poster submission are already open. Most obviously, given my current interests, I am quite excited by the prospect of taking part in this workshop (and sorry that Andrew can only take part by teleconference!).
Archive for Bayes factor
Quentin Gronau and Eric-Jan Wagenmakers just arXived a rather exotic paper in that it merges experimental mathematics with Bayesian inference. The mathematical question at stake here is whether or not one of the classical irrational constants like π, e or √2 are “normal”, that is, have the same frequency for all digits in their decimal expansion. This (still) is an open problem in mathematics. Indeed, the authors do not provide a definitive answer but instead run a Bayesian testing experiment on 100 million digits on π, ending up with a Bayes factor of 2×10³¹. The figure is massive, however one must account for the number of “observations” in the sample. (Which is not a statistical sample, strictly speaking.) While I do not think the argument will convince an algebraist (as the counterargument of knowing nothing about digits after the 10⁸th one is easy to formulate!), I am also uncertain of the relevance of this huge figure, as I am unable to justify a prior on the distribution of digits if the number is not normal. Since we do not even know whether there are
non-normal numbers outside rational numbers. While the flat Dirichlet prior is a uniform prior over the simplex, to assume that all possible probability repartitions are equally possible may not appeal to a mathematician, as far as I [do not] know! Furthermore, the multinomial model imposed on (at?) the series of digit of π does not have to agree with this “data” and discrepancies may as well be due to a poor sampling model as to an inappropriate prior. The data may more agree with H⁰ than with H¹ because the sampling model in H¹ is ill-suited. The paper also considers a second prior (or posterior prior) that I do not find particularly relevant.
For all I [do not] know, the huge value of the Bayes factor may be another avatar of the Lindley-Jeffreys paradox. In the sense of my interpretation of the phenomenon as a dilution of the prior mass over an unrealistically large space. Actually, the authors mention the paradox as well (p.5) but seemingly as a criticism of a frequentist approach. The picture above has its lower bound determined by a virtual dataset that produces a χ² statistic equal to the 95% χ² quantile. Dataset that stills produces a fairly high Bayes factor. (The discussion seems to assume that the Bayes factor is a one-to-one function of the χ² statistics, which is not correct I think. I wonder if exactly 95% of the sequence of Bayes factors stays within this band. There is no theoretical reason for this to happen of course.) Hence an illustration of the Lindley-Jeffreys paradox indeed, in its first interpretation of the clash between conclusions based on both paradigms. As a conclusion, I am thus not terribly convinced that this experiment supports the use of a Bayes factor for solving this normality hypothesis. Not that I support the alternative use of the p-value of course! As a sidenote, the pdf file I downloaded from arXiv has a slight bug that interacted badly with my printer in Warwick, as shown in the picture above.
I recently came across an ABC paper in PLoS ONE by Xavier Rubio-Campillo applying this simulation technique to the validation of some differential equation models linking force sizes and values for both sides. The dataset is made of battle casualties separated into four periods, from pike and musket to the American Civil War. The outcome is used to compute an ABC Bayes factor but it seems this computation is highly dependent on the tolerance threshold. With highly variable numerical values. The most favoured model includes some fatigue effect about the decreasing efficiency of armies along time. While the paper somehow reminded me of a most peculiar book, I have no idea on the depth of this analysis, namely on how relevant it is to model a battle through a two-dimensional system of differential equations, given the numerous factors involved in the matter…
Panayiota Touloupou (Warwick), Naif Alzahrani, Peter Neal, Simon Spencer (Warwick) and Trevelyan McKinley arXived a paper yesterday on Model comparison with missing data using MCMC and importance sampling, where they proposed an importance sampling strategy based on an early MCMC run to approximate the marginal likelihood a.k.a. the evidence. Another instance of estimating a constant. It is thus similar to our Frontier paper with Jean-Michel, as well as to the recent Pima Indian survey of James and Nicolas. The authors give the difficulty to calibrate reversible jump MCMC as the starting point to their research. The importance sampler they use is the natural choice of a Gaussian or t distribution centred at some estimate of θ and with covariance matrix associated with Fisher’s information. Or derived from the warmup MCMC run. The comparison between the different approximations to the evidence are done first over longitudinal epidemiological models. Involving 11 parameters in the example processed therein. The competitors to the 9 versions of importance samplers investigated in the paper are the raw harmonic mean [rather than our HPD truncated version], Chib’s, path sampling and RJMCMC [which does not make much sense when comparing two models]. But neither bridge sampling, nor nested sampling. Without any surprise (!) harmonic means do not converge to the right value, but more surprisingly Chib’s method happens to be less accurate than most importance solutions studied therein. It may be due to the fact that Chib’s approximation requires three MCMC runs and hence is quite costly. The fact that the mixture (or defensive) importance sampling [with 5% weight on the prior] did best begs for a comparison with bridge sampling, no? The difficulty with such study is obviously that the results only apply in the setting of the simulation, hence that e.g. another mixture importance sampler or Chib’s solution would behave differently in another model. In particular, it is hard to judge of the impact of the dimensions of the parameter and of the missing data.
Ritabrata Dutta, Malgorzata Bogdan and Jayanta Ghosh recently arXived a survey paper on model selection and multiple testing. Which provides a good opportunity to reflect upon traditional Bayesian approaches to model choice. And potential alternatives. On my way back from Madrid, where I got a bit distracted when flying over the South-West French coast, from Biarritz to Bordeaux. Spotting the lake of Hourtain, where I spent my military training month, 29 years ago!
“On the basis of comparison of AIC and BIC, we suggest tentatively that model selection rules should be used for the purpose for which they were introduced. If they are used for other problems, a fresh justification is desirable. In one case, justification may take the form of a consistency theorem, in the other some sort of oracle inequality. Both may be hard to prove. Then one should have substantial numerical assessment over many different examples.”
The authors quickly replace the Bayes factor with BIC, because it is typically consistent. In the comparison between AIC and BIC they mention the connundrum of defining a prior on a nested model from the prior on the nesting model, a problem that has not been properly solved in my opinion. The above quote with its call to a large simulation study reminded me of the paper by Arnold & Loeppky about running such studies through ecdfs. That I did not see as solving the issue. The authors also discuss DIC and Lasso, without making much of a connection between those, or with the above. And then reach the parametric empirical Bayes approach to model selection exemplified by Ed George’s and Don Foster’s 2000 paper. Which achieves asymptotic optimality for posterior prediction loss (p.9). And which unifies a wide range of model selection approaches.
A second part of the survey considers the large p setting, where BIC is not a good approximation to the Bayes factor (when testing whether or not all mean entries are zero). And recalls that there are priors ensuring consistency for the Bayes factor in this very [restrictive] case. Then, in Section 4, the authors move to what they call “cross-validatory Bayes factors”, also known as partial Bayes factors and pseudo-Bayes factors, where the data is split to (a) make the improper prior proper and (b) run the comparison or test on the remaining data. They also show the surprising result that, provided the fraction of the data used to proper-ise the prior does not converge to one, the X validated Bayes factor remains consistent [for the special case above]. The last part of the paper concentrates on multiple testing but is more tentative and conjecturing about convergence results, centring on the differences between full Bayes and empirical Bayes. Then the plane landed in Paris and I stopped my reading, not feeling differently about the topic than when the plane started from Madrid.
“Today, a week or two spent reading Jaynes’ book can be a life-changing experience.” (p.8)
I received this book by Peter Grindrod, Mathematical underpinnings of Analytics (theory and applications), from Oxford University Press, quite a while ago. (Not that long ago since the book got published in 2015.) As a book for review for CHANCE. And let it sit on my desk and in my travel bag for the same while as it was unclear to me that it was connected with Statistics and CHANCE. What is [are?!] analytics?! I did not find much of a definition of analytics when I at last opened the book, and even less mentions of statistics or machine-learning, but Wikipedia told me the following:
“Analytics is a multidimensional discipline. There is extensive use of mathematics and statistics, the use of descriptive techniques and predictive models to gain valuable knowledge from data—data analysis. The insights from data are used to recommend action or to guide decision making rooted in business context. Thus, analytics is not so much concerned with individual analyses or analysis steps, but with the entire methodology.”
Barring the absurdity of speaking of a “multidimensional discipline” [and even worse of linking with the mathematical notion of dimension!], this tells me analytics is a mix of data analysis and decision making. Hence relying on (some) statistics. Fine.
“Perhaps in ten years, time, the mathematics of behavioural analytics will be common place: every mathematics department will be doing some of it.”(p.10)
First, and to start with some positive words (!), a book that quotes both Friedrich Nietzsche and Patti Smith cannot get everything wrong! (Of course, including a most likely apocryphal quote from the now late Yogi Berra does not partake from this category!) Second, from a general perspective, I feel the book meanders its way through chapters towards a higher level of statistical consciousness, from graphs to clustering, to hidden Markov models, without precisely mentioning statistics or statistical model, while insisting very much upon Bayesian procedures and Bayesian thinking. Overall, I can relate to most items mentioned in Peter Grindrod’s book, but mostly by first reconstructing the notions behind. While I personally appreciate the distanced and often ironic tone of the book, reflecting upon the author’s experience in retail modelling, I am thus wondering at which audience Mathematical underpinnings of Analytics aims, for a practitioner would have a hard time jumping the gap between the concepts exposed therein and one’s practice, while a theoretician would require more formal and deeper entries on the topics broached by the book. I just doubt this entry will be enough to lead maths departments to adopt behavioural analytics as part of their curriculum… Continue reading
In this review paper, now published in Statistical Analysis and Data Mining 6, 3 (2013), David Parkinson and Andrew R. Liddle go over the (Bayesian) model selection and model averaging perspectives. Their argument in favour of model averaging is that model selection via Bayes factors may simply be too inconclusive to favour one model and only one model. While this is a correct perspective, this is about it for the theoretical background provided therein. The authors then move to the computational aspects and the first difficulty is their approximation (6) to the evidence
where they average the likelihood x prior terms over simulations from the posterior, which does not provide a valid (either unbiased or converging) approximation. They surprisingly fail to account for the huge statistical literature on evidence and Bayes factor approximation, incl. Chen, Shao and Ibrahim (2000). Which covers earlier developments like bridge sampling (Gelman and Meng, 1998).
As often the case in astrophysics, at least since 2007, the authors’ description of nested sampling drifts away from perceiving it as a regular Monte Carlo technique, with the same convergence speed n1/2 as other Monte Carlo techniques and the same dependence on dimension. It is certainly not the only simulation method where the produced “samples, as well as contributing to the evidence integral, can also be used as posterior samples.” The authors then move to “population Monte Carlo [which] is an adaptive form of importance sampling designed to give a good estimate of the evidence”, a particularly restrictive description of a generic adaptive importance sampling method (Cappé et al., 2004). The approximation of the evidence (9) based on PMC also seems invalid:
is missing the prior in the numerator. (The switch from θ in Section 3.1 to X in Section 3.4 is confusing.) Further, the sentence “PMC gives an unbiased estimator of the evidence in a very small number of such iterations” is misleading in that PMC is unbiased at each iteration. Reversible jump is not described at all (the supposedly higher efficiency of this algorithm is far from guaranteed when facing a small number of models, which is the case here, since the moves between models are governed by a random walk and the acceptance probabilities can be quite low).
The second quite unrelated part of the paper covers published applications in astrophysics. Unrelated because the three different methods exposed in the first part are not compared on the same dataset. Model averaging is obviously based on a computational device that explores the posteriors of the different models under comparison (or, rather, averaging), however no recommendation is found in the paper as to efficiently implement the averaging or anything of the kind. In conclusion, I thus find this review somehow anticlimactic.