Perform Jackknife-After-Bootstrap

DESCRIPTION:

Jackknife-after-bootstrap is a technique for estimating the standard error of some functional of the bootstrap distribution of parameter estimates. For example, it may be used to estimate standard errors for the bootstrap estimate of standard error. Jackknife-after-bootstrap also calculates relative influences reflecting the degree of influence each observation has upon the functional under consideration.

USAGE:

jack.after.bootstrap(boot.obj, functional=mean,  
                     threshold=2, ...,  
                     frame.eval.boot=sys.parent(1)) 

REQUIRED ARGUMENTS:

boot.obj
object of class bootstrap.

OPTIONAL ARGUMENTS:

functional
functional of the bootstrap distribution for which standard error is to be estimated. May be a character string specifying a column of boot.obj$stats (i.e. "Bias", "Mean", or "SE") or a function expecting a vector as its first argument and returning a scalar, e.g. mean.
threshold
observations with an absolute relative influence above this value will be flagged as particularly influential.
...
other arguments passed to functional.
frame.eval.boot
frame in which to evaluate the call to bootstrap used in resamp.get.indices to regenerate the indices used in constructing the replicates. By default, the function is evaluated in the frame which calls jack.after.boot.

VALUE:

object of class jack.after.boot with components call, functional, rel.influence , large.rel.influence, values.functional, dim.obs, and threshold.

WARNING:

Jackknife-after-bootstrap usually overestimates standard errors.

REFERENCES:

Efron, B. and Tibshirani, R. J. (1993). An Introduction to the Bootstrap. San Francisco: Chapman & Hall.

SEE ALSO:

, , .

EXAMPLES:

temp <- bootstrap(stack.loss,var) 
jack.after.bootstrap(temp)