From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Mon Apr 07 2003 - 12:16:24 PDT
On Mon, Apr 07, 2003 at 12:10:58PM -0700, Don Groom wrote:
> 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.
OK, I understood that-- but clearly I didn't make it clear in the
paper.
The other thing I never said is that they don't really do what they say
they do. By applying this prior, you end up with error bars (especially
for "too blue" supernovae) that are nowhere near Gaussian-- they're very
near half-Gaussian, yes, but not Gaussian. However, they then go on and
do a chisquare fit anyway. They *approximate* the new PDF for the color
uncertainty by using adjusting E(B-V) to the new mean value, and using
the second moment of the real PDF about this new distribution as the
error bar. This, anyway, is what I did when I tried to simulate their
methods, and it's what Peter swears to me that they do. I probably
should point out explicitly that I'm *not* really integrating the new
PDF you get as a result of applying the prior, but some (perhaps not
terribly good) approximation to it.
> *I like the discussion, although the mallet is still small. But then one
> of them might referee the paper.
Hmm-- I thought the mallet was pretty nasty already. I guess I'm just
oversensitive after glancing through Bob Kirshner's "mine is bigger than
yours" book... I'd like to avoid sinking to the same level of name
calling. (Not that I didn't just do that, but this is an internal
convrsation.)
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Mon Apr 07 2003 - 12:16:45 PDT