# boot: Bootstrap functions for LQM and LQMM In lqmm: 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

 ```1 2 3 4 5``` ```boot(object, R = 50, seed = round(runif(1, 1, 10000)), startQR = FALSE) ## S3 method for class 'lqm' boot(object, R = 50, seed = round(runif(1, 1, 10000)), startQR = FALSE) ## S3 method for class 'lqmm' 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.

Marco Geraci

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15``` ```# 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) ```

lqmm documentation built on April 13, 2018, 5:06 p.m.