4

I heard somewhere that in astrophysics, there are more bayesians than frequentists (although in general, I think that bayesian community is smaller). Is there something special about "astro" data what makes it more suitable for bayesian methods?

EDIT: The motivation for my question was also this kaggle competition, where the bayesian approach outperformed others. Therefore it seems that there could be something special about astro data.

sitems
  • 3,649
  • 1
  • 25
  • 52
  • 4
    How sure are you that the claim is true? – Glen_b Aug 05 '14 at 12:29
  • The first (about astrophysics) or the second one (general case)? – sitems Aug 05 '14 at 12:31
  • The first one.. – Glen_b Aug 05 '14 at 12:38
  • I do not know, but I heard it in a presentation by astrophysicist itself. – sitems Aug 05 '14 at 12:46
  • 6
    +1 I think this is a reasonable question, amenable to objective analysis, regardless of the truth of the preliminary claims it quotes. Indeed, there is a long history of astronomers being early adopters of statistical methods. Stephen Stigler has devoted considerable effort to addressing the question of why some fields (notably economics and psychology) historically took much longer to accept that statistical methods even had any relevance or application. – whuber Aug 05 '14 at 13:21

1 Answers1

2

One very common task in astronomy is image processing, concretely solving inverse problems which are under-determined. That is, given a distorted signal by an unknown process, estimate the original signal. A standard approach is to solve it though the maximum entropy principle (MaxEnt).

The paper by Steve Gull "Bayesian Inductive Inference and Maximum Entropy" explains it in great detail. Very nice paper. He was pioneering in popularizing those methods.

jpmuc
  • 12,986
  • 1
  • 34
  • 64