the case for TREC

Tom Anderson (tom@emigrant)
Wed, 17 Jun 1998 11:27:42 -0700 (PDT)

The rest of the story includes that they found it very
difficult to track down copies of the incorrect dataset,
leading to journal articles being published with old data
several years after they discovered the error. Hence TREC.
(For those not going to Amin's practice talk,
http://www.cs.berkeley.edu/WebOS/papers/trec.ps, to appear
at this week's USENIX.)

tom
---------
From: "Jeff Dozier" <dozier@bren.ucsb.edu>
To: "Tom Anderson" <tom@cs.washington.edu>, <vahdat@cs.washington.edu>
Subject: RE: Erroneous Satellite Data
Date: Sat, 13 Jun 1998 12:04:20 -0700
Message-ID: <000101bd96fe$12576960$c5eb47cf@owens.aol.silicon.com>
MIME-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
In-Reply-To: <199806130030.RAA21471@emigrant.cs.washington.edu>
Status: R

Amin/Tom:

I copied you on a message to Richard McPeters, who was the PI for the
Nimbus-7 TOMS (Total Ozone Mapping Spectrometer --
http://jwocky.gsfc.nasa.gov). The problem was that the early TOMS
measurements that showed the depleted ozone over Antarctica were correct,
but the program that analyzed them threw the values out and substituted
climatological estimates because they were so low. Later, when the depleted
ozone was noticed from surface measurements of ultraviolet radiation, people
went back and re-analyzed the old data.

Unfortunately the web site referenced above doesn't really tell this story.
It's often hard to convince scientists that we learn more from failure than
from success.

Jeff