qqnorm(x, method=<<see below>>, sig=.05, omit=F, slope, label="sig", datax=T, full=F, legend.loc, ...)
"fac.aov"
.
uses an estimate of the standard error of the factorial effects based on mean squared error.
"pse"
, unless
"mse"
is
available, in which case both
"pse"
and
"mse"
are used.
"sig"
, giving how many large effects to omit from plot.
If
omit
is an
integer, that many largest effects are omitted. If
omit = "sig"
all significant effects are omitted. A half-normal
plot of effects should "straighten out" when significant effects
are removed. By default, no effects are omitted from the plot.
TRUE
,
lines with intercept=0 and slope=1/(standard error) are plotted for
each method given in the method parameter (if
datax = F
, slope =
standard error). If
full = F
, significance limits are also plotted at
level given by
sig
.
"sig"
for labelling of points. If
FALSE
, no
labelling is done. If
TRUE
, the user is prompted to
identify the points to label. If a single number, n,
the n largest effects will be labeled. If
"sig"
, all of the
significant effects are
labeled, with significance determined by
sig
and
method
.
If more than one method is used, the method which
results in the most significant effects is used to label the
effects.
TRUE
, data goes on the
x-axis; if
FALSE
data goes on
the y-axis. Default is
TRUE
.
TRUE
produces a full-normal plot,
FALSE
a half-normal plot. When
full=TRUE
, the defaults become
slope=FALSE
, and
label=FALSE
. This creates a plot
which is useful for identifying possible outliers.
NULL
The underlying assumption of the normal and half-normal plot is that the inactive or unimportant effects will be normally distributed with zero mean and an unknown standard deviation. Thus, the line drawn on the normal or half-normal plot has zero intercept and slope equal to an estimate of the standard deviation of the "inactive" effects. In general, points on a half-normal plot falling off a line (i.e., below the line on the right) drawn through the "center" of the data are considered to be possibly important effects. This function will draw a line with intercept 0 and slope determined by one of several methods. Drawing the line on the plot is equivalent to estimating the standard error of the inactive effects. The following methods are available for drawing lines on the plot:
"mse"
: an estimate of the
standard error of the effects calculated from root mean squared error,
"pse"
: a psuedo standard
error estimate (Lenth, 1989, Haaland and O'Connell, 1994),
"tse"
: a 60% trimmed standard error of
the effects (Berk and Picard, 1991, Haaland and O'Connell, 1994), and
"ase"
: an adaptive estimate of the standard error of
the effects (Dong, 1993, Haaland and O'Connell, 1994).
Berk, K.N. and Picard, R.R. (1991). "Significance Tests for Saturated
Orthogonal Arrays." Journal of Quality Technology, 23, 174-178.
Box, G. E. P., W. G. Hunter, J. S. Hunter (1978),
Statistics for Experimenters,
New York: Wiley.
Daniel, C. (1959), Use of half-normal plots in interpreting factorial
two-level experiments, Technometrics, 1, 311-341.
Dong, F. (1993). "On the Idendification of Active Contrasts in
Unreplicated Fractional Factorials." Statistica Sinica, 3, 209-217.
Haaland, P. D. (1989),
Experimental Design in Biotechnology,
New York: Marcel Dekker.
Haaland, P. D. and M. A. O'Connell (1994), Inference
for effect saturated fractional factorials, to appear in Technometrics.
Lenth, R. V. (1989), Quick and easy analysis of unreplicated
factorials, Technometrics, 31, 469-473.
Nair, V. J. (1984), On the behavior of some estimators from
probability plots,
Journal of the American Statistical Association,
79, 823-831.
buffer.fac <- fac.aov(buffer.df) qqnorm(buffer.fac) # use all defaults qqnorm(buffer.fac, label=T) # interactively identify effects to label # specify different methods for plotting lines: qqnorm(buffer.fac, "ase", .10) qqnorm(buffer.fac, method="all") # omit certain significant effects from plot: qqnorm(buffer.fac, omit="sig", method="all") qqnorm(buffer.fac, omit=4, method="all") # put data on y axis: qqnorm(buffer.fac, datax=F) qqnorm(buffer.fac, datax=F, method="all") # produce full-normal plots with various options: qqnorm(buffer.fac, full=T) qqnorm(buffer.fac, full=T, datax=F, method="all")