Produces a quantile-quantile plot of the effects component of the fit against
the appropriate quantiles of the normal distribution.
USAGE:
qqnorm(x, full=F, label, omit=NULL, ...)
REQUIRED ARGUMENTS:
x:
the result of fitting an analysis of variance model, including the
cases of multivariate response and multiple-error strata.
OPTIONAL ARGUMENTS:
full:
TRUE for a full normal probability plot of the effects,
FALSE for a half-normal plot of the absolute values of the effects.
label:
should some of the points be labeled?
Can be set to TRUE, in which case the user will be prompted to identify
the points to label, or to a single number, n, in which case the n largest
effects (in absolute value) will be labeled.
omit:
those effects that should be omitted from the plot. By default, the first
effect (assumed to be the intercept) is omitted.
Either numeric or character data can be given.
Character data are matched against the effect
labels.
The intercept is always omitted: if you really want to see it, set its
name explicitly to anything other than "(Intercept)"---e.g.,
names(x$effects)[1] <- "*I*"
Graphical parameters may also be supplied as arguments
to this function (see par).
VALUE:
the points plotted, suitable to use as an argument to identify() to label
interesting points after the fact.
In the case of multiple responses or error strata, the points from the
last plot are returned.
SIDE EFFECTS:
A quantile-quantile plot is produced, either of the effects
component from the fit or
of its absolute value, plotted against the appropriate quantiles of the
normal distribution.
In the case of multiple response or multiple strata, one plot will
be produced for each separate set of effects.
Subtitle labels will be composed in this case identifying the response
and/or the stratum.
DETAILS:
This function is a method for the generic function
qqnorm
for class
"aov".
It can be invoked by calling
qqnorm
for an object of the appropriate class, or directly by calling
qqnorm.aov
regardless of the class of the object.