bootReg: Perform bootstrap computation on an object

In bozenne/butils: Various useful functions

Perform bootstrap computation on an object

Description

Perform bootstrap computation for a regression model. Handle one grouping variable.

Usage

bootReg(object, type, FUN.estimate, FUN.stdError, data, load.library, ...)

## S3 method for class 'lm'
bootReg(
  object,
  type = "coef",
  FUN.estimate = NULL,
  FUN.stdError = NULL,
  data = NULL,
  load.library = NULL,
  ...
)

## S3 method for class 'gls'
bootReg(
  object,
  type = "coef",
  FUN.estimate = NULL,
  FUN.stdError = NULL,
  data = NULL,
  load.library = "nlme",
  cluster,
  ...
)

## S3 method for class 'lme'
bootReg(
  object,
  type = "coef",
  FUN.estimate = NULL,
  FUN.stdError = NULL,
  data = NULL,
  load.library = "nlme",
  cluster,
  ...
)

## S3 method for class 'lvmfit'
bootReg(
  object,
  type = "coef",
  FUN.estimate = NULL,
  FUN.stdError = NULL,
  data = NULL,
  load.library = "lava",
  ...
)

.bootReg(
  object,
  data,
  strata = NULL,
  name.cluster,
  FUN.estimate,
  FUN.stdError,
  FUN.resample = NULL,
  FUN.iid = NULL,
  n.boot = 1000,
  n.cpus = 1,
  load.library,
  seed = 1,
  rejectIfWarning = TRUE,
  trace = TRUE
)

Arguments

object	the fitted model.
type	the type of test for which the bootstrap should be performed. Can be "coef", "anova", "publish". Setting type to NULL enable the use of FUN.estimate and FUN.stdError.
FUN.estimate	the function used to extract the punctual estimates from the model.
FUN.stdError	the function used to extract the standard error associated with the punctual estimate (i.e. standard error of the empirical estimator).
data	the data that have been used to fit the model.
load.library	additional library to load on each CPU. Useful when performing parallel computation.
...	ignored
cluster	the variable indicating the level where the sample is i.i.d. Only required for gls with no correlation argument.
strata	if not NULL, a stratified bootstrap is performed according to this variable.
name.cluster	internal argument.
FUN.resample	the function used simulate new data under the model. Default is NULL which corresponds to a non-parametric bootstrap.
FUN.iid	the function used to extract the influence function from the model.
n.boot	the number of replications. Should be a large number.
n.cpus	the number of cpu to use.
seed	set the random number generator
rejectIfWarning	Should the estimate be ignored if a warning is returned by the estimation routine?
trace	should the execution of the bootstrap be displayed using a progress bar?

Details

Bootstrap: randomly select observations (or individuals according to argument var.id) to form a new dataset. If the same individual appear several times, a different group value is given for each apparition.

When using multiple cores, even though a seed is set to each core, the result may change depending on how many samples each core is performing.

Examples

#### data  ####
n <- 1e2
set.seed(10)
df.data <- data.frame(Y = rnorm(n),
                     group = gl(3, 5, n, labels = c("Ctl","Trt","Neu")),
                     gender = gl(2, 5, n, labels = c("Female","Male"))[sample.int(n)]
                     )

#### lm ####
m.lm <- lm(Y ~ group*gender, data = df.data)
## Not run: 
resBoot <- bootReg(m.lm, n.boot = 1e4)

## End(Not run)

resBoot
summary(resBoot, type = "norm")
summary(resBoot, type = "basic")
summary(resBoot, type = "stud")
summary(resBoot, type = "perc")
summary(resBoot, type = "bca")

resBoot <- bootReg(m.lm, FUN.resample = "simulate", n.boot = 1e1)
resBoot

#### gls ####
library(nlme)
e.gls <- gls(follicles ~ sin(2*pi*Time) + cos(2*pi*Time),
             data = Ovary, correlation = corAR1(form = ~ 1 | Mare))
resBoot <- bootReg(e.gls, n.boot = 1e1)

#### lme ####
e.lme <- lme(follicles ~ sin(2*pi*Time) + cos(2*pi*Time),
             data = Ovary, random =~ 1 | Mare)
resBoot <- bootReg(e.lme, n.boot = 1e1)

bozenne/butils documentation built on Oct. 14, 2023, 6:19 a.m.