\HeaderA{boot.array}{Bootstrap Resampling Arrays}{boot.array}
\keyword{nonparametric}{boot.array}
\begin{Description}\relax
This function takes a bootstrap object calculated by one of the
functions \code{boot}, \code{censboot}, or \code{tilt.boot} and
returns the frequency (or index) array for the the bootstrap
resamples.
\end{Description}
\begin{Usage}
\begin{verbatim}
boot.array(boot.out, indices=)
\end{verbatim}
\end{Usage}
\begin{Arguments}
\begin{ldescription}
\item[\code{boot.out}] An object of class \code{"boot"} returned by one of the generation
functions for such an object.

\item[\code{indices}] A logical argument which specifies whether to return the frequency
array or the raw index array.  The default is \code{indices=FALSE}
unless \code{boot.out} was created by \code{tsboot} in which case the
default is \code{indices=TRUE}.

\end{ldescription}
\end{Arguments}
\begin{Details}\relax
The process by which the original index array was generated is
repeated with the same value of \code{.Random.seed}.  If the frequency
array is required then \code{freq.array} is called to convert the
index array to a frequency array.

A resampling array can only be returned when such a concept makes
sense.  In particular it cannot be found for any parametric or
model-based resampling schemes.  Hence for objects generated by
\code{censboot} the only resampling scheme for which such an array can
be found is ordinary case resampling. Similarly if \code{boot.out\$sim}
is \code{"parametric"} in the case of \code{boot} or \code{"model"} in
the case of \code{tsboot} the array cannot be found.  Note also that
for post-blackened bootstraps from \code{tsboot} the indices found
will relate to those prior to any post-blackening and so will not be
useful.

Frequency arrays are used in many post-bootstrap calculations such as
the jackknife-after-bootstrap and finding importance sampling weights.
They are also used to find empirical influence values through the
regression method.
\end{Details}
\begin{Value}
A matrix with \code{boot.out\$R} rows and \code{n} columns where
\code{n} is the number of observations in \code{boot.out\$data}.  If
\code{indices} is \code{FALSE} then this will give the frequency of
each of the original observations in each bootstrap resample. If
\code{indices} is \code{TRUE} it will give the indices of the
bootstrap resamples in the order in which they would have been passed
to the statistic.
\end{Value}
\begin{Section}{Side Effects}
This function temporarily resets \code{.Random.seed} to the value in
\code{boot.out\$seed} and then returns it to its original value at the
end of the function.
\end{Section}
\begin{SeeAlso}\relax
\code{\LinkA{boot}{boot}}, \code{\LinkA{censboot}{censboot}}, \code{\LinkA{freq.array}{freq.array}},
\code{\LinkA{tilt.boot}{tilt.boot}}, \code{\LinkA{tsboot}{tsboot}}
\end{SeeAlso}
\begin{Examples}
\begin{ExampleCode}
#  A frequency array for a nonparametric bootstrap
city.boot <- boot(city, corr, R=40, stype="w")
boot.array(city.boot)

perm.cor <- function(d,i) 
     cor(d$x,d$u[i])
city.perm <- boot(city, perm.cor, R=40, sim="permutation")
boot.array(city.perm, indices=TRUE)
\end{ExampleCode}
\end{Examples}