> Joe Hays said, among other things, "It is more of a challenge to
> "experiment" in my direct work with clients, who demand measurable
> progress and deliverables on schedule."
Rol Fessenden says:
> My company does not like
> experiments that fail, and yet, I am convinced that experimenting is
> crucial.
Ah, but what is "Failed Experiment"? Typically, managers know little
if anything about designed experiments, and especially sequential
Response Surface Designs. They were not exposed to these ideas in
their Management courses.
Once one sees the idea of a central composite design for a Response
Surface, it is clear that the prudent strategy is to move in steps
that assure that some of the results in the current design are
"worse" than the results in the last one. If all results are "worse"
that is good news if the design had enough power to detect important
differences. (The results have "boxed in" a local extremum.)
One interpretation of Rol's strategy of keeping several trials in
view is that each is a part of the larger sequential learning design.
Managers can benefit greatly from a half hour spent to understand the
process of a response surface design.
-----
Host's Note: OK, I'll bite. Tom, could you followup your own post and
tell us a little about "response surface design"? I expect that I'm not
the only one interested.
-----
-- Tom Johnson tom_johnson@ncsu.edu tel: (919) 515 4620 fax: (919) 515 1794 Box 8109 18A Patterson Hall North Carolina State University Raleigh, NC 27695-8109