From: "JACK L. URETSKY (C) 1996; HEP DIV., ARGONNE NATIONAL LAB, ARGONNE, IL 60439" <JLU@hep.anl.gov>
Date: Fri, 12 Dec 1997 22:38:21 -0600
Hi all-
Ludwik writes, boggling my mind:
***************************************************************
Before showing more data (see below) let me add another observation. The
rule which prevails in the lab should be that "data are bad only when
students do something incorrectly, otherwise they should be good". That
does not happen with "black boxes". During our experimentation we had to
reject data from numerous trials knowing very well that nothing was done
wrongly by us. Even in the case of coffee filters we often observed
strange curves and rejected them on the basis that "something is not
right" basis.
******************************
I hope that I'm misreading this. The burden, as I understand it,
is on the experimenter to demonstrate that the data are "good". The
ultimate test is that that the data are reproduced independently in
a different lab. Deviations from expectations should certainly be
investigated - consider the guy who kept moving his photographic film
away from the drawer containing the radium ore (before the discovery
or radioactivity).
In other words, the rejected data were just as "good" as the
accepted data. The rejection merely biased that experiment in favor of
your expectations.
Regards,
Jack