[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Error estimates addendum #2
- To: Peter Kammel <kammel@npl.uiuc.edu>
- Subject: Error estimates addendum #2
- From: Tom Banks <tbanks@socrates.berkeley.edu>
- Date: Tue, 27 Sep 2005 15:03:03 -0700 (PDT)
- Cc: analysis -- Tom Banks <tbanks@socrates.berkeley.edu>, Steve Clayton <smclayto@uiuc.edu>, Tim Gorringe <gorringe@pa.uky.edu>, Fred Gray <fegray@socrates.berkeley.edu>, David Hertzog <hertzog@uiuc.edu>, Brendan Kiburg <kiburg@npl.uiuc.edu>, Bernhard Lauss <lauss@socrates.berkeley.edu>, Francoise Mulhauser <Francoise.Mulhauser@psi.ch>, Claude Petitjean <Claude.Petitjean@psi.ch>, "R. Prieels" <prieels@fynu.ucl.ac.be>
- In-reply-to: <Pine.LNX.4.63.0509011522500.21119@one.npl.uiuc.edu>
- References: <Pine.LNX.4.63.0509011522500.21119@one.npl.uiuc.edu>
Hi all,
a little piece of good news regarding the uncertainties which I quoted in
today's teleconference: Fred and I were discussing the error formula
shortly afterwards, and Fred astutely noticed that all of the covariant
terms which I had ignored are actually *negative*. The [y2,c] error has
the explicitly negative coefficient (y1-y2), and the [lambda_b1,lambda_b2]
covariant errors will reduce things because the lambdas come from the same
data set--a semi-Kawall data set/subset effect.
Thus, the numbers I gave today are actually upper limits on the
uncertainty, and things are likely to improve once we figure out how to
properly deal with these covariant terms.
Regards,
Tom