At 6:40 AM -0700 11/2/2002, John S. Denker wrote:
You should have an analysis scheme in place, so that you can analyze the
data as it comes in, so that early results guide the design of later
measurements.
-------------------------------
Two comments on this thread.
1. I have never used DOE but I have seen it used and misused. I saw a
group of scientists at a national lab spend close to a year using this
approach to determine the effect (as an aside, many folks on this list
misuse the words effect and affect - and the effect of this misuse affects
me greatly (smiley face here)) of processing variables on the properties of
high temperature superconductors. They set up and ran their series of
experiments quite well and got some nice results. On the other hand, all
of us who were active in the field knew what the important processing
variables already were - they were already there in the literature - so an
expensive and time-consuming experiment reproduced what was already known.
This reminded me of the Feyman quote that just following a "scientific
process" is not necessarily science - it is merely following the forms of
science.
A good experiment is one that yields new and useful information that can be
used reproducibly in future efforts. The design of a good experiment is
the most useful skill that I learned in graduate school.
2. There is another aspect to experimental approach that is rarely taught,
yet is mentioned by John in his post above. That is, we often start at
point A and want to get to point Z - in developing new materials, processes
(fusion energy, solar energy, etc.), etc. The path from A to Z is not
clear although we have a sense of the general slope of how to get there.
We start at A and do a number of experiments B1, B2, B3, B4 that hopefully
will move us towards Z. We then find that B3 gives us results closer to Z
than the others. These results guide the design of the next phase of
experiments. So we now use B3 as our starting point and perform another
series of experiments C1, C2, C3, C4. We find that C2 provides a result
closer to our goal. This then continues until we reach our goal (or our
funding/project ends!).
Note that this approach often leads to patentable results. Whereas a
process using B3 may be obvious using A as a well known starting point, a
process using C2 may not be obvious, and so may be patentable.
Larry Woolf,General Atomics,San Diego, CA 92121; Phone: 858-526-8575; FAX:
858-455-8568
This posting is the position of the writer, not that of SUNY-BSC, NAU or the AAPT.