Finally, Richard Carrier’s blog is set up here, and now you can hear him explain Bayes’ Theorem.
People should start whining more now that “Skepticon” was a misnomer. It wasn’t a skeptics’ convention, it was a basic math and science convention!
Pharyngula
Evolution, development, and random biological ejaculations from a godless liberal
I already can’t keep up and the roster keeps growing. Not complaining about the addition, just lamenting my limits.
Dhorvath, OM #1
You’re not alone.
Well, I wasn’t there… so I can’t tell what happened.
Ken Ham said you can’t unless you witnessed it in person, and he should know. I mean, he wasn’t at the Creation 6700 years ago* (but some Sumerians were), he wasn’t at the Garden of Eden, but apparently some folks from the Land of Nod were* close by.
(*Creation date and Land of Nod purely speculative and depend entirely on your personal brand of magical thinking. your mileage may vary. void where prohibited by rationality )
Since the scientific method is rooted in skepticism, it makes sense that a skeptic’s conference would be teaching about skepticism’s most important tool.
Surely that’s not a bad thing, right?
Interesting post on the use of frequentist vs. Bayesian statistics in science:
http://oikosjournal.wordpress.com/2011/10/11/frequentist-vs-bayesian-statistics-resources-to-help-you-choose/
There’s far too much hyperbole for my liking including statements such as “if you are reasoning correctly then you are using Bayes’ Theorem” and “you can use Bayes’ Theorem to determine the cause”.
I’d also like to add that he’s waffling so much that
I can’t watch through to see if he can successfully explain Bayes’ Theorem (which means he failed to explain it to me). It’s actually very simple and most good books explain it in a short paragraph. I would recommend picking up a good introductory book on probabilities instead.
What? That statement is unjustified. Everyone
agreesknows that skepticism is an important aspect of doing science. But to say that skepticism is the root of science is taking it too far. And anyway, there are many facets which are equally important, such as curiosity and clear thinking.You may as well say that science is rooted in the love of God. For some, that might be true. But it is not a universal truth. Some people wrongly believe that Christianity made modern science possible (which Richard Carrier demonstrates to be false). Needless to say, it’s a little irritating that science gets hijacked by special interest groups.
Not really. Science is rooted in falsification of ideas. We try to prove our ideas false, which is skepticism in action. And by doing so, show the idea has merit if it can’t be refuted.
madscientist:
No, it means you failed to watch his explanation.
(It does mean he failed to hold your interest until he got to the specifics)
I think you have to have at least some skepticism with being a scientist, but I have encountered a number of insufficiently skeptical scientists in reading James Randi’s exploits.
(see Ref: Benveniste for example)
I don’t doubt the second sentence (which can be true for many endeavours which are not science).
But to say that science is rooted in the falsification of ideas… no, that is not science as I understand it (that being the investigation of nature). But skepticism is a function in scientific method, for sure.
Perhaps I don’t understand your meaning, though.
@6 torbertin wrote:
It might be an “interesting” post about frequentist vs. Bayesian statistics, but it isn’t a very enlightening one. Whenever I see an author write that there are “very deep philosophical differences” between Bayesian and frequentist statistics, I immediately know (with at least 0.95 probability) that they’re not very knowledgeable about either.
ah — so that’s how mr spock comes up with his annoying probabilities!
good to know.
Bah.
Show me your priors and I will present you with my posterior.
I mean, after all, what did RA Fisher know? Or Harold Jeffreys?
Also: The recent upswing in the use of Bayesian methods owes nothing to their inferential powers in comparison to frequentist methods. After all, inverse probability was in use long before frequentism.
The inclination toward “Bayesian” thinking is rather a response to improved computation which allows the estimation of the posterior distribution over complex parameter space. Given that most practitioners in the sciences* design priors to reflect ignorance, it seems clear that they do not have a truly “Bayesian” viewpoint, as much as they prefer a marginal likelihood estimator to a conditional one.
*Or at least in evolutionary biology
…as they should if there are many nuisance parameters to be estimated.
@17 A.E. wrote:
I doubt that. There is a grudgingly increasing awareness among scientists of the pitfalls of using p-values to test null hypotheses, judging by the number of recent critical journal articles in the scientific (as opposed to statistical) literature.
Are the papers to which you refer pro-Bayesian or just anti-p with Bayes as the default option?
In some applications (at least some of the ones that I work in), the pit-falls of formulating priors are much deeper and more difficult to cirmcumvent. Further, ML methods are often applied to systems in which no logical null exists (hence requiring no p) are statistically consistent, and unobjectionable from a Popperian standpoint.
“Unobjectionable from a Popperian standpoint”?
I was going to say—and I should have—that the other 5% were philosophers of statistics.
jt512: I’m not sure what you are getting at. In your opinion, is there is a real difference between Bayesian and frequentist statistics? If there is, and people are increasingly aware of it, do you mean to say that this increasing awareness is still limited to 5% of concerned scientists?
Of course there is a real difference between Bayesian and frequentist statistics. What I originally said is that when I read an article (and I meant on the internet) that makes a big deal about the philosophical differences, as the author of the blog I was referring to does, rather than the practical differences, that it usually means that the author doesn’t really understand either paradigm very well.
Contrast that blog post with an article like Wagenmakers [1]. In 21 pages of text (excluding references), Wagenmakers spends less time—two paragraphs—discussing the philosophical differences between Bayesian and frequentist statistics than did the blog author in a 9-paragraph article (excluding recommending readings). The remainder of Wagenmakers explains the deficiencies of the frequentist hypothesis test, the problems that it causes, and why and how Bayesian statistics is the better alternative.
As far as the percentage of “concerned” scientists that are aware of the difference (or the problem), I have no idea. The problem is the lack of concern, or awareness, among scientists. I’d say 5% awareness among scientists in general would be way too generous an estimate. However, I’m seeing an increasing frequency of articles appearing in the biomedical and experimental psychology literature like Wagenmakers’s. I think this, along with the occasional high-profile statistical disaster (eg, Bem [2]), is slowly leading to an increasing understanding of the limitations of frequentist statistics, and increasing use of Bayesian methods.
[1] Wagenmakers E-J. A practical solution to the pervasive problems of p values. Psychonomic Bulletin and Review 2007;14(5); 779–804.
[2] Bem, DJ. Feeling the Future: Experimental evidence for anomalous retroactive influences on cognition and affect. Journal of Personality and Social Psychology 2011; 100; 407-425.