Rq: rms Package Interface to quantreg Package

Description

The Rq function is the rms front-end to the quantreg package's rq function. print and latex methods are also provided, and a fitting function RqFit is defined for use in bootstrapping, etc. Its result is a function definition.

For the print method, format of output is controlled by the user previously running options(prType="lang") where lang is "plain" (the default), "latex", or "html". For the latex method, html will actually be used of options(prType='html').

Usage

Rq(formula, tau = 0.5, data=environment(formula),
   subset, weights, na.action=na.delete,
   method = "br", model = FALSE, contrasts = NULL,
   se = "nid", hs = TRUE, x = FALSE, y = FALSE, ...)
# S3 method for Rq
print(x, digits=4, coefs=TRUE, title, ...)
# S3 method for Rq
latex(object,
           file = '', append=FALSE,
           which, varnames, columns=65, inline=FALSE, caption=NULL, ...)
# S3 method for Rq
predict(object, ..., kint=1, se.fit=FALSE)
RqFit(fit, wallow=TRUE, passdots=FALSE)

Value

Rq returns a list of class "rms", "lassorq" or "scadrq", "Rq", and "rq". RqFit returns a function definition. latex.Rq returns an object of class "latex".

Arguments

formula: model formula
tau: the single quantile to estimate. Unlike rq you cannot estimate more than one quantile at one model fitting.
data,subset,weights,na.action,method,model,contrasts,se,hs: see rq
x: set to TRUE to store the design matrix with the fit. For print is an Rq object.
y: set to TRUE to store the response vector with the fit
...: other arguments passed to one of the rq fitting routines. For latex.Rq these are optional arguments passed to latexrms. Ignored for print.Rq. For predict.Rq this is usually just a newdata argument.
digits: number of significant digits used in formatting results in print.Rq.
coefs: specify coefs=FALSE to suppress printing the table of model coefficients, standard errors, etc. Specify coefs=n to print only the first n regression coefficients in the model.
title: a character string title to be passed to prModFit
object: an object created by Rq
file,append,which,varnames,columns,inline,caption: see latexrms
kint: ignored
se.fit: set to TRUE to obtain standard errors of predicted quantiles
fit: an object created by Rq
wallow: set to TRUE if weights are allowed in the current context.
passdots: set to TRUE if ... may be passed to the fitter

Author

Frank Harrell

Examples

Run this code

if (FALSE) {
set.seed(1)
n <- 100
x1 <- rnorm(n)
y <- exp(x1 + rnorm(n)/4)
dd <- datadist(x1); options(datadist='dd')
fq2 <- Rq(y ~ pol(x1,2))
anova(fq2)
fq3 <- Rq(y ~ pol(x1,2), tau=.75)
anova(fq3)
pq2 <- Predict(fq2, x1)
pq3 <- Predict(fq3, x1)
p <- rbind(Median=pq2, Q3=pq3)
plot(p, ~ x1 | .set.)
# For superpositioning, with true curves superimposed
a <- function(x, y, ...) {
 x <- unique(x)
 col <- trellis.par.get('superpose.line')$col
 llines(x, exp(x), col=col[1], lty=2)
 llines(x, exp(x + qnorm(.75)/4), col=col[2], lty=2)
}
plot(p, addpanel=a)
}