car (version 2.0-21)

showLabels: Utility Functions to Identify and Mark Extreme Points in a 2D Plot.

Description

This function is called by several graphical functions in the car package to mark extreme points in a 2D plot. Although the user is unlikely to call this function directly, the documentation below applies to all these other functions.

Usage

showLabels(x, y, labels=NULL, id.method="identify",  
  id.n = length(x), id.cex=1, id.col=palette()[1], ...)

Arguments

x

Plotted horizontal coordinates.

y

Plotted vertical coordinates.

labels

Plotting labels. If NULL, case numbers will be used. If labels are long, the substr or abbreviate function can be used to shorten them.

id.method

How points are to be identified. See Details below.

id.n

Number of points to be identified. If set to zero, no points are identified.

id.cex

Controls the size of the plotted labels. The default is 1.

id.col

Controls the color of the plotted labels.

...

additional arguments passed to identify or to text.

Value

• A utility function primarily used for its side-effect of drawing labels on a plot. Returns invisibly the labels of the selected points, or NULL if no points are selected. Although intended for use with other functions in the car package, this function can be used directly.

Details

The argument id.method determine how the points to be identified are selected. For the default value of id.method="identify", the identify function is used to identify points interactively using the mouse. Up to id.n points can be identified, so if id.n=0, which is the default in many functions in the car package, then no point identification is done. Automatic point identification can be done depending on the value of the argument id.method.

• id.method = "x"select points according to their value ofabs(x - mean(x))
• id.method = "y"select points according to their value ofabs(y - mean(y))
• id.method = "r"select points according to their value ofabs(y), as may be appropriate in residual plots, or others with a meaningful origin at 0
• id.method = "mahal"Treat(x, y)as if it were a bivariate sample, and select cases according to their Mahalanobis distance from(mean(x), mean(y))
• id.methodcan be a vector of the same length asxconsisting of values to determine the points to be labeled. For example, for a linear modelm, settingid.method=cooks.distance(m), id.n=4will label the points corresponding to the four largest values of Cook's distance, orid.method = which(abs(residuals(m, type="pearson")) > 2would label all observations with Pearson residuals greater than 2 in absolute value. Warning: If missing data are present, points may be incorrectly labelled.
• id.methodcan be a vector of case numbers or case-labels, in which case those cases will be labeled. Warning: If missing data are present, a list of case numbers may identify the wrong points. A list of case labels, however, will work correctly with missing values.

With showLabels, the id.method argument can be a list, so, for example id.method=list("x", "y") would label according to the horizontal and vertical axes variables. Finally, if the axes in the graph are logged, the function uses logged-variables where appropriate.

References

Fox, J. and Weisberg, S. (2011) An R Companion to Applied Regression, Second Edition, Sage. Weisberg, S. (2014) Applied Linear Regression, Fourth Edition, Wiley.

See Also

avPlots, residualPlots, crPlots, leveragePlots

Examples

plot(income ~ education, Prestige)
with(Prestige, showLabels(education, income, 
     labels = rownames(Prestige), id.method=list("x", "y"), id.n=3))
m <- lm(income ~ education, Prestige)
plot(income ~ education, Prestige)
abline(m)
with(Prestige, showLabels(education, income, 
     labels=rownames(Prestige), id.method=abs(residuals(m)), id.n=4))