Makes a half or full normal plot for the effects from an aov model

Makes a half or full normal plot for the effects from a model inheriting from class aov. One can interactively label the points in the plot.

# S3 method for class 'aov'
qqnorm(y, full=FALSE, label=FALSE, omit=NULL,
           xlab=paste(if (full) "" else "Half", " Normal plot"), 
           ylab="Effects", ...)

Arguments

y

A model object inheriting from aov

full

Full or half normal plot (half is default)

label

If TRUE, function allows interactive labelling of points in plot, using the mouse

omit

Numeric or character vector of effects to omit, the intercept is always omitted

xlab

Horizontal axix label

ylab

Vertical axis label

...

Further arguments to be given to the plot function

Details

Produces a (half) normal plot of the effects from an AOV model. The idea behind the plot is that most effects will be small or null, and this effects can be used as a basis for estimation of the experimental variance. This small effects will show up in the plot as a straight line, other effects can be judged against this as a background. Heavily used by Box, Hunter & Hunter, which attributes the idea to Daniel.

Value

If label=TRUE, the vector of points identified, else nothing of interest.

References

Box, Hunter and Hunter: Statistics for Experimenters. An Introduction to Design, Data Analysis and Model Building. Wiley.
Daniel, C (1976): Applications of Statistics to Industrial Experimentation. Wiley.
Daniel, C (1959): Use of half-normal plot in interpreting factorial two-level experiments. Technometrics.1, 149.

Author

Kjetil Halvorsen kjetil@entelnet.bo

Examples

library(MASS)
data(npk)
npk.aov <- aov(yield ~ block + N*P*K, npk)
qqnorm(npk.aov)


## interactive labeling of points.  Click mouse on points to show label.
if (dev.interactive()) qqnorm(npk.aov, omit=2:6, label=TRUE)