From: Don Groom (deg@panisse.lbl.gov)
Date: Mon Apr 07 2003 - 12:10:58 PDT
Rob et al.,
Just a comment about the discussion of the R98 prior.* A Bayesian
prior (a little redundant) can be anything one wishes, and not all are
necessarily biased. In fact, the main objection I and others have with
Bayesian methods is that the answer depends on the prior. In the case of
R98 the prior is a one-sided Gaussian, and I think this should be said.
Other priors might be E(B-V)=0 (the right side of Fig 8 uses this one),
a symmetric Gaussian about 0 (which describes the experimental data if
outliers are discarded), and a skewed distribution which describes the
experimental distribution including a probability for reddened objects.
The latter, which does include negative experimental values, might or
might not be best, but so little is known about the distribution that it
would become just another arbitrary prior.
The whole point is that R98 used a biased prior resulting from their
confusion of true value with measured value, their fundamental error.
But it's a particular choice of prior, not just a prior.
D
*I like the discussion, although the mallet is still small. But then one
of them might referee the paper.
|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|
Don Groom (Particle Data Group, Supernova Cosmology Project)
DEGroom(at)lbl.gov www-ccd.lbl.gov Voice: 510/486-6788 FAX: 510/486-4799
Analog: 50-308//Berkeley Lab//Berkeley, CA 94720
This archive was generated by hypermail 2.1.4 : Mon Apr 07 2003 - 12:11:31 PDT