rqmcmb {rqmcmb2} | R Documentation |
MCMB for Quantile Regression (also see quantreg package by Roger Koenker)
rqmcmb(x=x, y=y, tau=0.5, K=100, int=TRUE, plotTheta=FALSE)
x |
a data matrix (n by p) for the design variables whose rows correspond to cases |
y |
a response vector of length n |
tau |
a percentile level between 0 and 1. Default at 0.5 for the median |
K |
length of the MCMB sequence. Default is 100 |
plotTheta |
TRUE or FALSE for plotting the MCMB sequence. Default to FALSE |
int |
should be set to TRUE if the intercept is to be included in the model, and to FALSE if no intercept is desired. Default is TRUE. |
A list with the following components:
coef |
the parameter estimate from rq() |
theta |
a matrix containing the MCMB sequence. The first row is the initial parameter estimate from rq() |
success |
returns 1 if MCMB is successful. A value of 0 indicates that the program fails to return a desired MCMB sequence |
cn |
condition number of the X'X matrix. |
The MCMB may not be suitable for problems of small sample sizes. Severe collinearity in the x matrix could also be harmful.
Maria Kocherginsky (mkocherg@health.bsd.uchicago.edu) and Xuming He (x-he@uiuc.edu)
Kocherginsky, M., He, X. Extensions of the Markov Chain Marginal Bootstrap. Statistics & Probability Letters, in press.
Kocherginsky M, He X, Mu Y. (2005). Practical confidence intervals for regression quantiles. Journal of Computational and Graphical Statistics 14:41-55, 2005.
Kocherginsky, M. (2003). Extensions of the Markov Chain Marginal Bootstrap. Ph.D Thesis, University of Illinois Urbana-Champaign.
He, X. and Hu, F. (2002). Markov Chain Marginal Bootstrap. Journal of the American Statistical Association , Vol. 97, no. 459, 783-795.
library(quantreg) x <- cbind(rnorm(100), runif(100)) y <- rnorm(100) #generate the MCMB sequence: mcmb <- rqmcmb(x, y, tau=.5, plotTheta=FALSE) #get MCMB estimates of mean, SD, and CI: rqmcmb.ci(mcmb) #plot the MCMB sequences: rqmcmb.plot(mcmb)