A popular and economical approach that gives information only
on the effects of single factors, but not on interactions, is the
Plackett–Burman (PB) method, introduced in 1946 when the
authors were working for the British Ministry of Supply. This
method is well suited to ruggedness testing, i.e. establishing
whether the outcome of an analytical procedure is affected by
changes in each relevant factor. The most important feature of
PB designs is that they all involve 4n experiments, where n ¼ 1,
2, 3. In each case the maximum number of factors that can be
studied is 4n 1, so an 8-experiment PB design can study no
more than 7 factors, a 12-experiment design will handle up to 11
factors, and so on. This may seem to be inconvenient, but it
turns out to be a valuable feature of the method.
Suppose we wish to study four factors. Four experiments will
be then insufficient, so we shall have to use eight experiments in
a PB design, and have seven factors. This means that three oftimes at the higher level and four times at the lower. The
effect of each factor is then readily determined from the
expression:
2[P(y+) P(y)]/N
where N is the total number of experiments, eight in this case.
The (y+) terms are the responses when a given factor is at its
high level, and the (y) terms reect the responses for that
factor set to its low level. It can be shown that the effects for the
main factors determined in this way are not confounded with
each other (see AMCTB 36).
A popular and economical approach that gives information onlyon the effects of single factors, but not on interactions, is thePlackett–Burman (PB) method, introduced in 1946 when theauthors were working for the British Ministry of Supply. Thismethod is well suited to ruggedness testing, i.e. establishingwhether the outcome of an analytical procedure is affected bychanges in each relevant factor. The most important feature ofPB designs is that they all involve 4n experiments, where n ¼ 1,2, 3. In each case the maximum number of factors that can bestudied is 4n 1, so an 8-experiment PB design can study nomore than 7 factors, a 12-experiment design will handle up to 11factors, and so on. This may seem to be inconvenient, but itturns out to be a valuable feature of the method.Suppose we wish to study four factors. Four experiments willbe then insufficient, so we shall have to use eight experiments ina PB design, and have seven factors. This means that three oftimes at the higher level and four times at the lower. Theeffect of each factor is then readily determined from theexpression:2[P(y+) P(y)]/Nwhere N is the total number of experiments, eight in this case.The (y+) terms are the responses when a given factor is at itshigh level, and the (y) terms reect the responses for thatfactor set to its low level. It can be shown that the effects for themain factors determined in this way are not confounded witheach other (see AMCTB 36).
การแปล กรุณารอสักครู่..
A popular and economical approach that gives information only
on the effects of single factors, but not on interactions, is the
Plackett–Burman (PB) method, introduced in 1946 when the
authors were working for the British Ministry of Supply. This
method is well suited to ruggedness testing, i.e. establishing
whether the outcome of an analytical procedure is affected by
changes in each relevant factor. The most important feature of
PB designs is that they all involve 4n experiments, where n ¼ 1,
2, 3. In each case the maximum number of factors that can be
studied is 4n 1, so an 8-experiment PB design can study no
more than 7 factors, a 12-experiment design will handle up to 11
factors, and so on. This may seem to be inconvenient, but it
turns out to be a valuable feature of the method.
Suppose we wish to study four factors. Four experiments will
be then insufficient, so we shall have to use eight experiments in
a PB design, and have seven factors. This means that three oftimes at the higher level and four times at the lower. The
effect of each factor is then readily determined from the
expression:
2[P(y+) P(y)]/N
where N is the total number of experiments, eight in this case.
The (y+) terms are the responses when a given factor is at its
high level, and the (y) terms reect the responses for that
factor set to its low level. It can be shown that the effects for the
main factors determined in this way are not confounded with
each other (see AMCTB 36).
การแปล กรุณารอสักครู่..