This document is divided into a number of sections:
* High-level resampling functions
* General help files (information for a variety of resampling areas)
* Nonparametric Bootstrap
* Permutation Tests
* Jackknife, Influence, and ABC Limits
* Parametric and Smoothed Bootstrap
* Parametric Bootstrap Test
* Cross Validation
* Bootstrap Prediction Errors
* General-purpose S-PLUS functions
* Graphical Interface (Windows only)
The functions that perform resampling are:
describes various problems you may have, and describes remedies where possible.
contains details on arguments to arguments used by many high-level resampling functions.
describes various options.
The initial routines are and . The latter bootstraps the difference in statistics between two samples.
More details on arguments, see . For different types of samplers, see: ; note that any sampler can be combined with stratfied/multi-group sampling (the "group" argument to bootstrap and other functions) and sampling by subject (the "subject" argument)
Bootstrap and other objects: , , . Other model objects are handled without special methods.
Print, summarize, plot: , , , , ,
Description of a "bootstrap" object, extract parts: , , , .
Diagnostics: , .
Confidence intervals: , , , , .
Modify a "bootstrap" object: , , .
Control variate adjustment to improve accuracy: , .
Utility routines: and other routines in the same help file, , .
The initial routines are , , and . The latter two are specifically for comparing the differences between two samples, for general statistics or a mean (or column means), respectively.
More details on arguments, see . Note that calls , so many of the arguments are common.
Combination of p-values for multivariate statistics, or across groups in the case of : and other functions in the same help file.
Print, summarize, plot: , , , ,
Description of a "permutationTest" object, extract parts: , , , .
Modify a "permutationTest" object: .
The initial routines are: , , and
More details on many arguments, see .
Jackknife and other objects: , , . Other model objects are handled without special methods.
Print, summarize, plot: , , , , .
Description of these objects, extract parts: , , , , .
Confidence intervals: , .
Modify a "jackknife" or "influence" object: .
Utility routines: and other routines in the same help file, .
The initial routines are and for parametric and smoothed bootstrapping, respectively. The latter is implemented by calling the former; while smoothed bootstrapping seems at first glance to be more similar to simple nonparametric bootstrapping, it is closer to parametric bootstrapping in terms of the confidence intervals and other quantities that can be calculated from the results.
More details on arguments, see .
Print, summarize, plot: , , , , , , ,
Description of a "parametricBootstrap" or "smoothedBootstrap" object, extract parts: , .
Confidence intervals: , .
Modify a "parametricBootstrap" or "smoothedBootstrap" object: , .
Utility routines: and other routines in the same help file.
The initial routine is .
More details on some arguments, see .
Print, summarize, plot: , , , ,
Description of a "parametricBootstrapTest" object, extract parts: , .
Modify a "parametricBootstrapTest" object: , .
Utility routines: and other routines in the same help file.
The initial routines are , , and ,
More details on some arguments, see .
Print: ,
Modify a "crossValidation" object: .
The initial routines are , , and ,
More details on some arguments, see .
Print: ,
Modify a "bootstrapValidation" object: .
There are a number of replacements for existing S-PLUS functions. The following functions add a "weights" argument: , , , , , , , , , , , , , .
The following add other new options, or fix bugs: , , , , .
The following general-purpose functions are new:
: random sample with no or minimal replacement
,
,
,
: these are more general than their name suggests; use them to simultaneously select subsets of data and compute summaries.
,
: cumulative distribution function, and plot
: tabulate each column. Supports weights -- this provides a fast way to compute sums of a variable by levels of a factor.
: return all combinations of "k" numbers out of "n".
,
,
,
: calculations and random generation for discrete distributions.
,
,
: saddlepoint estimates for the mean of observations from a discrete distribution (supports stratified sampling).
,
,
,
: summaries, by a single grouping variable. Optionally replicate the summaries and place them in the order of the grouping variable (e.g. useful for subtracting mean from each group).
: subtract the mean for each group.
: inverse of a function, solve nonlinear equation.
: fit monotone interpolating curve through data.
: simple random sample without replacement, from a vector, matrix, or data frame
(an enhanced
),
, , : saddlepoint calculations, for sample means or sums of data; stratification and weights are allowed.
The following modeling functions and utilities are modified to support resampling: , , , , and .
There is a graphic interface for the S+Resample package available on Windows. This includes supplements to existing Spotfire S+ menus for some common operations, and menus for using core resampling capabilites (bootstrap, jackknife, and permutation tests) with your choice of statistics. For help on the graphical interface, start by using the menu `Help:Available Help:Resample'.