ebplot(x, sig=.05, method=<<see below>>,
n.effects= <<see below>>, ...)
"fac.aov", from fitting
fractional factorial model using
fac.aov.
n.effects is
not provided.
method="mse" uses root mean squared
error,
"pse" uses a pseudo standard error estimate,
"tse" uses
a 60% trimmed standard error of the effects.
"ase" uses an
adaptive standard error estimate. If
"mse" is available, it is
the default, otherwise,
"pse" is the default.
method is used to determine how many
effects are significant.
acplot.
ebplot calls
acplot after computing empirical Bayes estimates of
alpha
, the proportion of active effects, and
k, the variance of
the active effects.
If
neffects is specified, based on knowledge from previous analysis
graphical displays,
alpha is
neffects/length(feffects). If
n.effects
is not specified, significant effects are identified from
the test based on
sig and
method.
alpha is then estimated as the
fraction of significant effects. The scale factor,
k, is estimated
by the square root of the F-value from the least squares model
containing only the significant effects.
Box and Meyer (1986), An analysis for unreplicated fractional factorials, Technometrics, 28, 11-18.
buffer.fac <- fac.aov(buffer.df) ebplot(buffer.fac) # use all defaults # specify 2 as the number of effects thought to be significant: ebplot(buffer.fac,n.effects=2) # use estimated standard errors to identify significant effects: ebplot(buffer.fac,.05,'ase') ebplot(buffer.fac,.2,'tse')