Jeremy Fox often publishes interesting blogposts like today’s. I’m “reblogging” straight from his site as an experiment.
Why ecologists might want to read more philosophy of science
Categories: Error Statistics
12 Comments
12 thoughts on “Why ecologists might want to read more philosophy of science”
I welcome constructive comments that are of relevance to the post and the discussion, and discourage detours into irrelevant topics, however interesting, or unconstructive declarations that "you (or they) are just all wrong". If you want to correct or remove a comment, send me an e-mail. If readers have already replied to the comment, you may be asked to replace it to retain comprehension.
The Statistics Wars & Their Casualties
LSE PH500 Research Seminar (May 21-June 25, 2020): Controversies in Phil Stat
Summer Seminar 2019 (article)
Experts convene to explore new philosophy of statistics field
Error Statistics Philosophy 
They didn’t carry over the cute fox pic. You’ll have to go to the site to see that.
So my father is an ecologist and I wish he read less philosophy of science. The other day we were having dinner and I said something like:
“You should really try Bayesian statistics! It is really the most flexible approach, you can assume whichever probability distributions you want, it would be a much better fit for the kind of models and data you’re working with.”
And he (basically) replies:
“The Bayesian approach is no good because Deborah Mayo says so.”
For some reason he’s been reading you and come to this conclusion (which is strange because as I read you you argue against a Bayesian philosophy of statistics, not against doing actual Bayesian statistics, right?).
Frequentist vs Bayesian is a philosophical question as long as it only has philosophical implications. But when my father reads philosophy of science and comes to the conclusion that it is not kosher/halal to fit a hierarchical model in BUGS/JAGS/STAN then Frequentist vs Bayesian is not just a philosophical question anymore. If reading philosophy of science stops ecologists from running usefull statistcal analyses using usefull statistical software then perhaps it would be better to read less philosophy of science. Or perhaps read some more usefull philosophy of science at least…
Wha wha wha!? Mayo, please tell this ecologist that multilevel models are error-statistically kosher; only the methodology for M-S testing and primary inference is under dispute, and there are well-established frequentist approaches for these in multilevel models! (rasmusab, tell your dad to look into the econometric literature on panel data.)
I like your father!
But seriously, I’m madly dashing to finish final exam cheat sheets for my critical thinking class–will return.
Rasmusab:
Getting back to you and your father, you wrote: “For some reason he’s been reading you and come to this conclusion (which is strange because as I read you you argue against a Bayesian philosophy of statistics, not against doing actual Bayesian statistics, right?)”.
If I thought one’s philosophy of statistics had nothing to do with the warrant for doing actual statistics in accord with that philosophy, I wouldn’t dream of working on philosophy of statistics. So I’m not sure what you mean by the above. Possibly, you’re reacting to something I say in EGEK (Mayo 1996), to the effect that my focus (there!) had more to do with the use of Bayesian ideas in philosophy of science than Bayesian practice. That’s because my focus there was solving recalcitrant problems of philosophy of science (e.g.,problem of induction, underdetermination, Duhem’s problem,nature of science vs pseudoscience). You see, I feel that Bayesian logics of evidential-relationship (E-R logics), like confirmation theories and Bayesian formal epistemologies, reflect philosopher’s having given up on solving those problems (in the 60s-70s). Instead they “rationally reconstruct” things like “confirmation” and “explanation” appealing either to subjective “epistemic” probabilities, or else they allude largely to events in simple games of chance.
For a quick comparison of error statistics as a statistical philosophy and as a philosophy of statistics please see:
https://errorstatistics.com/2012/11/30/error-statistics-brief-overview/
But elsewhere, and certainly after EGEK, I have talked about statistical philosophy (which does refer to actual statistics).
Irony and bad faith: https://errorstatistics.com/2011/12/11/irony-and-bad-faith-deconstructing-bayesians-1/
Statistical science meets philosophy of science:
https://errorstatistics.com/2012/12/02/statistical-science-meets-philosophy-of-science/
I’ve been lecturing critical thinking students for some hours this evening, so this is rushed…
Bottom line: maybe father knows best.
“maybe father knows best.”
Not even my mother has ever said that….
Dear Mayo,
I am a quantitative ecologist, and I have been following your blog for some time now. I am happy to see this reference to philosophy of science by an ecologist.
Likewise, I think that reading about philosophy of statistics helps me getting a more global perspective about the underlying principles and questions that make the field of statistics evolve.
I was essentially formed as a frequentist but I have become interested in Bayesian ideas quite recently. Reading your error statistics philosophy has not deterred me from studying more about Bayesian ideas. But it surely helped me to better understand the reasoning beyond hypothesis testing.
Hope you could suggest me, as an ecologist (not a philosopher or a mathematical statistician) a similar synthesis of the Bayesian philosophical perspective. I quite enjoyed reading the following introductory book, Stone JV (2013) Bayes’ Rule. A tutorial introduction to Bayesian analysis. Sebtel Press, 170 pp., but would like to go further.
All the best
Luís Silva
Of course one of the best known philosophical problems of confirmation concerns (armchair) ornithology!
Jeremy Fox: I need to be honest regarding something you say in your article:
“One thing philosophers (again, at least in the “analytic” tradition) are good at is having productive arguments. In my experience, you don’t see philosophers just repeating the same points and ignoring the arguments of their opponents, hoping to win the argument by sheer force of repetition.”
Philosophers may pride themselves on doing this, and often they do—but an exception seems to be philosophy of confirmation/statistics. That is why my post of April 1, 2011 could even be an “April Fool’s” joke. https://errorstatistics.com/2012/04/01/3102/
Some people were hopeful when they first saw this post, until I deflated them by pointing to the date. I don’t mean people adhere to a given Bayesian howler in the face of a frequentist clarification/response, I mean that the clarification/response is never mentioned, not once, and texts such as Howson and Urbach repeat the howler verbatim, as do the next generation of students. I’m prepared to admit/allow the possibility that this would be much less likely to occur were I a male.
Good point. And a depressing one, on two counts. As an outsider to philosophy, I like my idealized image of it as a field in which there’s a “culture” of productive debate. It’s depressing to be told that my idealized mental image isn’t even a near approximation to reality, at least when it comes to philosophy of statistics. And it’s doubly depressing to be told that sexism might be the reason (not surprising, of course–I may idealize philosophy a bit but I’m not naive–but depressing all the same).
Given that some of the debates in which you’ve been engaged are long-standing ones that have been quite polarized for a while, do you think it’s possible that some of the difficulty you’ve had in getting Bayesians to engage with your arguments is down to that long-standing polarization? Not that that’s mutually exclusive with sexism, of course.
Jeremy: I want to make 100% sure that a vague, speculative conjecture of mine (which I’m sure wouldn’t wholly explain things) isn’t taken as an allegation with actual evidence. A hunch or feeling, to me, is not evidence in the least. So I had better not answer your leading questions.
Pingback: Foreword; To Tomorrow | The New Antiquated Taxi Dog Blues