Forums  > Books & Papers  > Looking for "Why I am not a Bayesian" by Clark Glymour  
     
Page 1 of 1
Display using:  

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-08 14:17

I'd be very grateful if anybody could send me a copy of this paper. Failing that, any other paper that sets out a well-argued case against the Bayesian approach.

Thanks.


heartbreakingly simple

M1kz


Total Posts: 121
Joined: Sep 2004
 
Posted: 2006-12-08 14:49
This?

The hardest thing to explain is the glaringly evident which everybody had decided not to see.

Holmes


Total Posts: 185
Joined: Dec 2004
 
Posted: 2006-12-08 14:54

When I had to do this I used this page

http://www-cdf.fnal.gov/physics/statistics/statistics_recommendations.html

 

They had some papers 'arguments' for and against

I emphasise 'had'. I know Jack Shit about this now-a-days.


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-08 15:51

M1kz, the exact one. Thanks.

Holmes - a useful page. Thanks to you too.

 


heartbreakingly simple

saffron


Total Posts: 181
Joined: May 2006
 
Posted: 2006-12-08 16:09

I remember reading a while back on Wikipedia about bayesian vs frequentist approaches to probability. Perhaps you can look there (and follow some links therein). Look for "bayesian probability" and "frequency probability".

 


jungle
Chief Rhythm Officer
CSD LLC
Total Posts: 3163
Joined: Jul 2004
 
Posted: 2006-12-08 17:24

"The anti-Bayesian arguments which convinced me — and a very good book in general — are those of Mayo (1996)..." (cosma shalizi)

amazon link

 


"buy at 90 degrees, sell at 270 degrees"

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-08 17:41
Thanks Jungle. My tendencies being skepticism and empiricism (rather than God, Truth and Descartes) I'm naturally inclined to being a Bayesian. So I'd like to read a well-argued case for the opposition. Your suggestion looks just the thing.

heartbreakingly simple

nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2006-12-09 01:13
a 2-sentence summary of why some people don't like bayesian-ness (although annoyingly math-free) is in this recent essay in scientific american

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-09 12:25

Do you mean the use of priors? The frequentist approach interprets probabilities as frequencies of occurrences given infinite populations. The dependency of posterior probabilities upon choice of prior goes to zero after "infinite" occurrences, so Bayesian and frequentist approaches both give the same results under frequentist assumptions. It's a trivial non-issue.

For the avoidance of doubt, I'm not looking for references to Wikipedia or pop-science articles. I'm looking for well-argued technical material by authors with a strong grasp not only of statistical theory but also of philosophy of science. There are many strong proponents of the Bayesian approach (Cox, Jayne, etc). I'm looking for papers that put forward a coherent anti-Bayesian view.


heartbreakingly simple

saffron


Total Posts: 181
Joined: May 2006
 
Posted: 2006-12-09 16:37

I'm sorry, I accidently wrote links when I was refering to references. I wouldn't call some of the references in those wikip articles "pop-science articles". I just thought your problem with the bayesian approach was the use of priors and not say the probabilistic approach to handling data. That's why I thought the debate with the frequentist approach may expose some of the arguements against the bayesian approach.

Anyway, I just thought this may assist you, apparently it doesn't so, sorry to have wasted your time.

Cheers.

 

 


IAmEric
Phorgy Phynance
Banned
Total Posts: 2961
Joined: Oct 2004
 
Posted: 2006-12-09 18:10
Have you seen this?

Our brain is a Bayesian computer

Researchers at the University of Rochester may have answered one of neuroscience's most vexing questions—how can it be that our neurons, which are responsible for our crystal-clear thoughts, seem to fire in utterly random ways?


I know nothing about the subject, but thought it was interesting when I stumbled onto it when it came out.

crowlogic
Banned

Total Posts: 270
Joined: Jan 2006
 
Posted: 2006-12-11 20:46
I'm not sure how anyone could be "anti-Bayesian", it is simple logic and deduction. Now, many things people call "Bayesian" I am certainly against. One of the main arguments against I've heard is about the arbitrary choice of prior, and I certainly agree. Just because something is "Bayesian" doesnt make it good, if you choose a shitty prior then your model will be shitty. Conjugate priors are interesting but very limited, there are some concepts related to this, so called 'quantum bayesian' math, etc.

One should respect public opinion insofar as is necessary to avoid starvation and keep out of prison, but anything that goes beyond this is voluntary submission to an unnecessary tyranny. --Bertrand Russell

nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2006-12-19 14:30
here is a very recent article called "The Case for Objective Bayesian Analysis" (by a bayesian) which discusses anti-bayesian critiques.


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-19 14:35
Thanks Nuno. Looks like an interesting paper. I'll read it in the next few days.

heartbreakingly simple

nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2006-12-20 21:29
in case you're still interested, the brillian breiman had something to say in the bayesian-bashing in this 1997 nontechnical letter "no bayesians in foxholes"

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-12-21 13:02
Excellent - thanks!

heartbreakingly simple

Holmes


Total Posts: 185
Joined: Dec 2004
 
Posted: 2006-12-21 13:47

I can vaguely remember some practical  'argument/anecdote': it went along the lines that quite often you are making a prior assumption whether you are aware of it or not.

Imagine you are trying to measure a partciles mass from its momentum. You get some readings that tell you the mass is negative. You use that data. Now you notice that prior or not you have assumed that mass can be negative. For some new particle you might go hmmm but if its a snooker ball...

Sooooo be a good scientist before you are a good statistician


nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2006-12-21 14:04
Holmes wrote


Imagine you are trying to measure a partciles mass from its momentum. You get some readings that tell you the mass is negative. You use that data. Now you notice that prior or not you have assumed that mass can be negative. For some new particle you might go hmmm but if its a snooker ball...


one can definitely think of situations in which having a prior makes all the difference, and in which it's almost impossible to think about the situation without a prior. one of my favorite interview brainteasers is the two envelopes problem.

Holmes


Total Posts: 185
Joined: Dec 2004
 
Posted: 2006-12-21 17:37
edit: BTW I am neither a good scientist or a good statistician

nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2007-02-10 07:20
and yet still more: "in class bayesian demos" (from a bayesian):

http://www.stat.columbia.edu/~gelman/research/published/bayesdemos.pdf
Previous Thread :: Next Thread 
Page 1 of 1