boot: Bootstrap functions for LQM and LQMM

In gforge/lqmm_gforge: Linear Quantile Mixed Models

Description

This function is used to obtain a bootstrap sample of a fitted LQM or LQMM. It is a generic function.

Usage

boot(object, R = 50, seed = round(runif(1, 1, 10000)), startQR = FALSE)

Arguments

object	an object of class "lqm" or "lqmm".
R	number of bootstrap replications.
seed	optional random number generator seed.
startQR	logical flag. If TRUE the estimated parameters in object are used as starting values in the algorithm applied to each bootstrap sample. This may cause the algorithm to converge too often to a similar optimum, which would ultimately result in underestimated standard errors. If FALSE (recommended), starting values are based on lm.

Value

An object of class boot.lqm is a data frame with R rows and npars columns containing the bootstrap estimates of theta. If object contains results for multiple quantiles, boot.lqm returns an array of dimension c(R,npars,nt), where nt is the length of tau.

An object of class boot.lqmm is a data frame with R rows and npars columns containing the bootstrap estimates of theta_x, theta_z, and scale. If object contains results for multiple quantiles, boot.lqmm returns an array of dimension c(R,npars,nt), where nt is the length of tau. The elements of theta_z are labelled with reStruct. See function covHandling and the example below on how to derive the variance-covariance matrix of the random effects starting from theta_z.

The following attributes are available:

tau	index of the quantile(s).
estimated	the estimated parameter as given by object.
R	number of bootstrap replications.
seed	the random number generator seed used to produce the bootstrap sample.
npars	total numer of parameters.
rdf	the number of residual degrees of freedom.
indices	the bootstrap sample of independent data units.

Author(s)

Marco Geraci

Examples

# boot.lqm
set.seed(123)
n <- 500
test <- data.frame(x = runif(n,0,1))
test$y <- 30 + test$x + rnorm(n)
fit.lqm <- lqm(y ~ x, data = test, tau = 0.5)
fit.boot <- boot(fit.lqm)
str(fit.boot)

# boot.lqmm
data(Orthodont)
fit <- lqmm(distance ~ age, random = ~ 1, group = Subject,
	tau = 0.5, data = Orthodont)
fit.boot <- boot(fit)
str(fit.boot)

gforge/lqmm_gforge documentation built on Dec. 20, 2021, 10:42 a.m.