rsm.design(n.factors, factor.names.arg = std.factor.names(n.factors),type = "cc", n.cp = min( n.factors + 1, 4), alpha, logx = F, fraction)
type
is
"cc"
,
"fc"
or
"bb"
, each element of the list gives two numbers, specifying the
"cube" points of each factor, i.e., the "Low" and "High"
points ( See details below.). The default levels are
c(-1,1)
for
natural scale and
c(.01, 10)
for
log scale. If
type
is
"3^n"
, each element gives
three numbers, specifying the three levels of each factor.
The default levels are
c(-1,0,1)
for natural scale and
c(.01, 1, 10)
for
log scale.
"cc"
= central composite,
"fc"
= face centered cube,
"3^n"
=
factorial design with three levels per factor, and
"bb"
=
Box Behnkens design.
min(number of factors + 1,4)
.
2^(nc/4)
, where nc is the number of points in the cube, or fractional
factorial part of the design.
logx
, has
length 1, the same value is used for all factors.
rsm.design
which inherits from
design
and
data.frame
. Each of the factors in the
returned design has class
rsm.factor
. See documentation of
rsm.factor
object
for a discussion of their attributes.
A central composite design has three components: the "cube" (a
fractional factorial design on the Low and High points), center points
(in the center of the cube), and star points (points extended from
the center of each face of the cube).
Each factor in the returned design has five values:
Low Star
,
Low
,
Center
,
High
and
High Star
.
Low
and
High
are given in the
factor.names.arg
list, if supplied. They default to -1 and 1.
Center
is the average of
Low
and
High
.
Low Star
=
Center - alpha * (High -Low) * 0.5
;
Hi Star
=
Center - alpha * (Hi - Low) * 0.5
A face centered design is a central composite design with
alpha = 1
.
This means each factor only has three levels, rather than the five
of the general central composite design.
The factorial design with three levels per factor, is a full
factorial design, and typically requires a large number of runs (
3^n
,
where n is the number of factors)
Box-Behnken designs are designs with only three levels per factor
that are economical in the number of runs required. In S+DOX, these are
available for up to 7 factors.
Box, G.E.P. and Draper, N.R (1987),
Empirical Model Building and Response
Surfaces, New York: Wiley.
Haaland, P.D. (1989),
Experimental Design in Biotechnology,
New York: Marcel Dekker.
# generate a central composite design rsm.design(2,list(RadDos=c(100,300),Prime1=c(7,21))) # generate a central composite design with factors on # a log scale rsm.design(2,list(A=c(2,8),B=c(1,100)),log=c(T,T)) # generate a central composite design which has as its # base fraction a resolution IV design rather than the # default resolution V design rsm.design(7, fraction = 1/8) # generate a factorial design with 3 factors each at 3 levels rsm.design(3, n.cp = 3, type= "3^n") # generate a Box-Behnken design with the first two # factors on a log scale rsm.design(6, type = "bb", log = c(T,T,F,F,F,F)) # generate a face centered design rsm.design(3,alpha=1)