“StatSci meets PhilSci”. (1/14/14)
As “Wasserman on Wasserman” (and links within) continues to rack up record hits (N.D.: you see you shouldn’t quit blogging)*, I’ve been asked about the origins of his and related discussions on this blog. For a quick answer:** many grew out of attempts to tackle the general question: “Statistical Science and Philosophy of Science: Where Do (Should) They meet?”–the title of a conference I organized (with A. Spanos***) at the London School of Economics, Center for the Philosophy of Natural and Social Science, CPNSS, in June 2010. In tackling this question, errorstatistics.com regularly returns to a set of contributions stemming from the conference, and conversations initiated soon after (with Andrew Gelman and Larry Wasserman)****. The conference site is here. My reflections in this general arena (Sept. 26, 2012) are here.
Opening with an informal (recorded) exchange: “A statistical scientist meets a philosopher of science: a conversation between Sir David Cox and Deborah Mayo”, this special topic of the on-line journal, Rationality, Markets and Morals (RMM), edited by Max Albert[i],—also a conference participant —has been an excellent home for continual updates (to which we may return at some point!)
Authors are: David Cox, Andrew Gelman, David F. Hendry, Deborah G. Mayo, Stephen Senn, Aris Spanos, Jan Sprenger, Larry Wasserman
To those who ask me what to read as background to some of the issues, have a look at those contributions. Many of them are discussed in specific blogposts (with “deconstructions” [by me], responses by authors, and insightful “U-Phil” analyses by readers) and comments.[ii]. (Search under U-Phil.) I have gathered a list of issues that we either haven’t taken up, or need to return to.
Here is the RMM blub:
Rationality, Markets and Morals: Studies at the Intersection of Philosophy and Economics
Guest Editors: Deborah G. Mayo, Aris Spanos and Kent W. Staley
Statistical Science Meets Philosophy of Science: The Two-Way Street
At one level of analysis, statisticians and philosophers of science ask many of the same questions: What should be observed and what may justifiably be inferred from the resulting data? How well-tested or confirmed are hypotheses with data? How can statistical models and methods bridge the gaps between data and scientific claims of interest? These general questions are entwined with long standing philosophical debates, so it is no wonder that the statistics crosses over so often into philosophical territory.
The “meeting grounds” of statistical science and philosophy of science are or should be connected by a two-way street: while general philosophical questions about evidence and inference bear on statistical questions (about methods to use, and how to interpret them), statistical methods bear on philosophical problems about inference and knowledge. As interesting as this two-way street has been over many years, we seem to be in need of some entirely new traffic patterns! That is the basis for this forum.
[i] Along with Hartmut Kliemt and Bernd Lahno.
[ii] The “deconstruction” activity on this blog began with my reaction to a paper by Jim Berger, in a recently reblogged post. Berger had replied in ‘Jim Berger on Jim Berger’.
*From the WordPress 2013 “annual report”: The busiest day of the year was February 18th. The most popular post that day was R. A. Fisher: how an outsider revolutionized statistics.
Also attracting huge hits was the guest post by Larry Laudan: Why Presuming Innocence is Not a Bayesian Prior: https://errorstatistics.com/2013/07/20/guest-post-larry-laudan-why-presuming-innocence-is-not-a-bayesian-prior/ Many other biggies (especially from guest posters) have attracted a large number of comments and views.
** This post adapts an earlier one here. This blog is on philosophy, after all: only careful and frequent rereading brings illumination.
***For a full list of collaborators, sponsors, logisticians, and related collaborations, see the conference page. The full list of speakers is found there as well. Should we do a 2015 update? or wait for ERROR 2016?
****Conference participants who never got around to sending papers: I think there’s still time.
Jose Bernardo was at our conference, and his presentation, he said, was taken from the 9th Valencia meeting just before ours. I was looking at it yesterday because of a surprising quote I came across in a paper by high energy particle physicist Robert Cousins. Bernardo’s paper is here:
Click to access bernardo-integrated-objective-bayesian-estimation-and-hypothesis-testing.pdf
The quote is cited below. Lindley, one of the commentators on Bernardo’s Valencia paper, is duly shocked that not only does Bernardo violate the likelihood principle (as he’s done in the past) but also denigrates the popular conception that p-values exaggerate the evidence against the null! (Bernardo is a Lindley student.) Bernardo is prepared to chuck Bayes factors.
In the discussion, p. 59, Bernardo writes:
“Thus, I certainly do not agree with [a commentator] that Lindley’s paradox has been ‘misunderstood’ as an illness of Bayes factors for precise hypothesis testing. On the contrary, this clearly poses a very serious problem to Bayes factors, in that, under certain conditions, they may lead to misleading answers. Whether you call
this a paradox or a disagreement, the fact that the Bayes factor for the null may be arbitrarily large for sufficiently large n, however relatively unlikely the data may be under H0 is, to say the least, deeply disturbing.”
I’ve only scanned the paper, but this seems a departure, and the commentators make it sound as if its Bernardo’s “last of Valencia” shocker (well not quite). Does Bernardo’s move show that by convenient choice of “loss function” the Bayesian can get essentially the same answers as the frequentist? Maybe I convinced him (that the significance tester gets the right answer), but I am convinced now that it was surely the particle physicists (at something called PhyStat). Anyway, I wanted to park his paper here, and see what people thought.