rbentfit: rank estimation for bent line regression

In Rbent: Robust Bent Line Regression

Description

This function use Wilcoxon score functions for fitting the bent line regression model.

Usage

rbentfit(y, z, x, bet.ini, tau.ini, tol = 1e-04, max.iter = 50)

Arguments

y	A vector of response
z	A vector of covariates
x	A numeric variable with change point
bet.ini	A initial vector of regression coefficients
tau.ini	A initial value of change point
tol	tolerance value, 1e-4 for default
max.iter	the maximum iteration steps

Value

A list with the elements

est	The estimated regression coefficients with intercept.
bp	The estimated change point.
est.se	The estimated standard error of the regression coefficients.
bp.est	The estimated standard error of the change point.
iter	The iteration steps.

Author(s)

Feipeng Zhang

Examples

n <- 150
x <- runif(n, 0, 4)
z <- rnorm(n, 1, 1)
y <- 1 + 0.5*z + 1.5*x  - 3 *pmax(x-2, 0)  + rt(n, 2)
rbentfit(y, cbind(1,z), x, bet.ini = c(0, 1, 1, -2), tau.ini = 1)

# for the example of  MRS data
data(data_mrs)
x <- log(data_mrs$mass)
y <- log(data_mrs$speed)
z <- data_mrs$hopper
tau.ini <- 3
dat.new <- data.frame(y=y, z1=z, z2 = x, z3=pmax(x-tau.ini,0))
library(Rfit)
fit.ini <- rfit(y~ z1 + z2 +z3, data= dat.new)   # with intercept
bet.ini <- fit.ini$coef
fit.rank <- rbentfit(y, cbind(1,z), x, bet.ini, tau.ini)

Rbent documentation built on May 2, 2019, 8:25 a.m.