Unformatted text preview: probably fail. Though a
statistician can play this role, it is often better played
by another member of the experimentation team,
who can “champion” the experiment among peers.
The statistician (or surrogate) can play a strong support role and be primarily responsible for that in
which he or she is professionally trained-the
and analysis of the experiment and not the execution.
Finally, considering item 13 of the guide sheet, the
team should entertain the idea of trial runs to precede
if this is the first in a
series of experiments. A trial run can consist of a
centerpoint run or a small part (perhaps a block) of
the experiment. The first and most important purpose of trial runs is to learn and refine experimental
procedures without risking the loss of time and expensive experimental samples. Most experiments involve people (and sometimes machines) doing things
that they have never done before. Usually some practice helps.
A second important reason for trial runs is to estimate experimental error before expending major
resources. An unanticipated large experimental error could lead to canceling or redesigning the experiment, widening the ranges of settings, increasing
the number of replicates, or refining the experimental procedure. An unanticipated small experimental
error (does this ever really happen?) could have opposite effects on plans or cause the experimenters to
reassess whether the estimate is right or complete.
A third reason is that trial runs are also excellent
opportunities to ensure that data-acquisition systems
are functioning and will permit experimental runs to
be conducted as fast as had been planned.
Last, a fourth reason is that trial runs may yield
results so unexpected that the experimenters decide
to change their experimental plans.
Naturally, the feasibility and advisability of conducting trial runs depends on the context, but the
experiment teams in which we have been involved
have never regretted conducting trial runs. Some trial
runs have saved experiments from disaster.
10. SUMMARY To...
View Full Document
- Fall '14
- The Land, response variable, Technometrics, guide sheet, david e. coleman