jackknife
.
jackknife(data, statistic, args.stat=NULL, seed=0, group.size=1, assign.frame1=F)
args.stat
. If it is an expression and the
data
argument is the
name of an object, then the object should also be referred to by name in the
expression. If the data is constructed within the call to
jackknife
, the
data should be referred to as
data
in the expression. (See help for
bootstrap
for examples of usage.)
statistic
when calculating
the statistic on the resamples.
group.size
is not equal to one.
May be a legal random
number seed or an integer between 0 and 1000 which will be passed to
set.seed
.
group.size=1
, the standard delete-1 jackknife is performed.
Otherwise, the
observations are divided into
floor(n/group.size)
groups of equal size
and these groups are jackknifed. Although this is similar to delete-d
jackknifing, all possible subsets of the specified size are not used, and
the jackknife statistics treat the replicates as a standard jackknife sample
of size
floor(n/group.size)
. This is provided primarily to allow grouped
jackknifing when calculating acceleration for BCa confidence intervals.
jackknife
estimates are identical, try setting
assign.frame1=T
. Note that this
will slow down the algorithm.
jackknife
which inherits from
resamp
. This has
components
call
,
observed
,
replicates
,
estimate
,
B
,
n
,
dim.obs
,
and
seed.start
. The data frame
estimate
has
three columns containing the bootstrap estimates of
Bias
,
Mean
, and
SE
.
assign.frame1=T
, the user must be sure that this assignment does not
overwrite some quantity of interest stored in frame 1.
Performs nonparametric jackknifing of observations for a wide scope of statistics and expressions.
Efron, B. and Tibshirani, R. J. (1993).
An Introduction to the Bootstrap.
San Francisco: Chapman & Hall.
Shao, J. and Tu, D. (1995).
The Jackknife and Bootstrap.
New York: Springer-Verlag.
jackknife(stack.loss,var) # See help(bootstrap) for more examples of syntax.