lmfor (version 1.2)

mywhiskers: A whiskers type residual plot

Description

A function for adding vertical lines onto residual plots to show

1. 95% confidence intervals of means or

2. 95% confidence intervals for individual observations

in the classes of the variable on the x-axis. Plot of the first type is useful for analyzing the fit of the assumed fixed part and plots of type b can be used to analyze the homogeneity of residuals.

Usage

mywhiskers(x, y, 
           nclass = 10, 
           limits = NA, 
           add = FALSE, 
           se = TRUE, 
           main = "", 
           xlab = "x", 
           ylab = "y", 
           ylim = NA, 
           lwd = 1, 
           highlight = "red")

Arguments

x

The variable on the x-axis. Usually one of the predictors or the predicted value.

y

The variable on the y-axis. Usually model residual.

nclass

The maximum number of classes to be used.

limits

The class limits. Alternative to nclass.

add

logical. Whether a new graphic window is opened or the lines will be added into an exosting plot.

se

Logical. Use standard errors of means (se=TRUE, option (a) above) or class-specific standard deviations (se=FALSE, option (b) above).

main, xlab, ylab, ylim, lwd

Graphical parameters of the plot. ignored if add=TRUE.

highlight

The color for lines that do not cross the y-axis.

Value

The function is usually used for its side effects (i.e., the plot). However, the values used in producing the plot are returned in a list of elements

• x: the class middlepoint x values.

• m: class-specific means of y.

• s: class-specific standard deviations or standard errors of y (see details).

• lb: lower ends of the class-specific lines.

• ub: upper ends of the lines.

Details

The function first classifies the data in nclass classes of variable x so that each class has approximately equal number of observations. Then the class mean and deviation s is computed for each class, where s is either the standard error of the mean (if se=TRUE) or standard deviation (if se=FALSE). A vertical line is plotted at the middle of each class showing the class mean by a dot and lines of length 3.92*s. If the line does not cross the x- axis, then the highlight color is used in the line. With small number of observations (or lot of ties), the number of classes is decreased until each class includes the minimum of 2 observations.

References

Mehtatalo, L. 2010. Forest biometrics with R. Lecture notes.

Examples

# NOT RUN {
x<-seq(1,100,1)
y<-x+10*log(x)+rnorm(100,0,5)
fm1<-lm(y~x)
plot(x,resid(fm1))
mywhiskers(x,resid(fm1),se=FALSE,add=TRUE)
mywhiskers(x,resid(fm1),se=TRUE,lwd=2,add=TRUE)
abline(h=0)
# }