Plot Jackknife-After-Bootstrap Diagnostics

DESCRIPTION:

Plot either: (1) The functional (e.g. quantiles, for each leave-one-out bootstrap distribution) vs. any of jackknife statistics, influence values, or the original data. (2) Standardized influence of each observation on the functional, analogous to a Cook's distance plot.

USAGE:

plot.jackknifeAfterBootstrap(x, nrow=NULL, grid.layout=T, id.outliers=T, 
   threshold=x$threshold, ..., graphical = x$graphical, 
   xaxis, superpose, Func, subset.x, subset.plots, absolute = F) 

REQUIRED ARGUMENTS:

x
object of class .

OPTIONAL ARGUMENTS:

nrow
number of rows of plots on the screen. If NULL and quantiles are not being plotted and x$dim.func is non-NULL, the first dimension is used; otherwise this is set to two if multiple plots are needed.
grid.layout
logical flag indicating whether to display plots in a single grid or as separate plots when the resampled statistic is a vector.
id.outliers
logical flag indicating whether observation numbers should be printed for standardized influences greater than the specified threshold.
threshold
threshold at which points are considered to have a relatively large influence. A straight line is drawn at this level, and observations with a standardized influence greater than this threshold are labeled if id.outliers=T.
...
additional arguments may be passed on to plot to alter the default appearance of the plot.
graphical
logical, if TRUE then the functional is plotted; if FALSE then standardized influence is plotted. Default is the value of x$graphical.
xaxis
what values to use on the x axis, one of "Observation" (observation number, default for influence plot), "L" (empirical influence function) "jackknife" - (jackknife statistics = leave-one-out statistics, default for plot of functional), "data" - original data.
superpose
logical, if TRUE then columns of the function are combined on a single plot; if FALSE they are shown on separate plots. By default this is TRUE when the functional is known to consist of quantiles.
Func
logical, if TRUE (the default) then when the functional is plotted, horizontal dotted lines are included at the values of the functional for the original bootstrap distribution.
subset.x
vector of integers or names; if xaxis is one of "L", "jackknife", or "data", you may use this to select a column for use in plotting.
subset.plots
integer vector, use this to do only a subset of the plots that would otherwise be produced.
absolute
logical, if TRUE then absolute values of standardized influence are plotted; if FALSE then negative values are shown.

VALUE:

no value is returned.

SIDE EFFECTS:

For each column (if the functional is multivariate), a plot is produced in which high-density lines are used to display the standardized influences. A solid horizontal line is plotted at the value of threshold. If id.outliers=T, then values greater than threshold are labeled with their observation number.

SEE ALSO:

, .

EXAMPLES:

x <- qexp(ppoints(19)) 
boot <- bootstrap(x, mean) 
plot(boot) # slightly skewed 
jab <- jackknifeAfterBootstrap(boot) 
plot(jab) 
plot(jab, xaxis = "L") 
plot(jab, xaxis = "data") 
 
jab <- jackknifeAfterBootstrap(boot, functional = "Centered Quantiles") 
plot(jab) 
 
# See also extensive examples in the help file for jackknifeAfterBootstrap