These functions can be used to construct standard errors, confidence intervals and tests of hypotheses regarding quantile regression models.
boot.rq(x, y, tau = 0.5, R = 200, bsmethod = "xy", mofn = length(y),
coef = NULL, blbn = NULL, cluster = NULL, U = NULL, ...)A list consisting of two elements:
A matrix B of dimension R by p is returned with the R resampled
estimates of the vector of quantile regression parameters. When
mofn < n for the "xy" method this matrix has been deflated by
the factor sqrt(m/n).
A matrix U of sampled indices (for bsmethod in c("xy", "wxy"))
or gradient evaluations (for bsmethod in c("pwy", "cluster"))
used to generate the bootstrapped realization, and potentially reused
for other taus when invoked from summary.rqs.
The regression design matrix
The regression response vector
The quantile of interest
The number of bootstrap replications
The method to be employed. There are (as yet) five options: method = "xy" uses the xy-pair method, and method = "pwy" uses the method of Parzen, Wei and Ying (1994) method = "mcmb" uses the Markov chain marginal bootstrap of He and Hu (2002) and Kocherginsky, He and Mu (2003). The "mcmb" method isn't compatible with sparse X matrices. The fourth method = "wxy" uses the generalized bootstrap of Bose and Chatterjee (2003) with unit exponential weights, see also Chamberlain and Imbens (2003). The fifth method "wild" uses the wild bootstrap method proposed by Feng, He and Hu (2011).
optional argument for the bootstrap method "xy" that permits subsampling (m out of n) bootstrap. Obviously mofn should be substantially larger than the column dimension of x, and should be less than the sample size.
coefficients from initial fitted object
orginal sample size for the BLB model
If non-NULL this argument should specify cluster id
numbers for each observation, in which case the clustered version of
the bootstrap based on the proposal of Hagemann (2017). If present
bsmethod is set to set to "cluster". If this option is used
and the fitting method for the original call was "sfn" then the
bootstrapping will be carried out with the "sfn" as well. This
is usually substantially quicker than the older version which
employed the "br" variant of the simplex method. Use of "sfn"
also applies to the "pwy" method when the original fitting
was done with "sfn". Finally, if na.action = "omit" and
length(object$na.action) > 0 then these elements are also
removed from the cluster variable. Consequently, the
length of the cluster variable should always be the same
as the length of the original response variable before any
na.action takes place.
If non-NULL this argument should specify an array of indices
or gradient evaluations to be used by the corresponding bootstrap
method as specified by bsmethod. This is NOT intended as
a user specified input, instead it is specified in summary.rqs
to ensure that bootstrap samples for multiple taus use the same
realizations of the random sampling.
Optional arguments to control bootstrapping
Roger Koenker (and Xuming He and M. Kocherginsky for the mcmb code)
Their are several refinements that are still unimplemented. Percentile methods should be incorporated, and extensions of the methods to be used in anova.rq should be made. And more flexibility about what algorithm is used would also be good.
[1] Koenker, R. W. (1994). Confidence Intervals for regression quantiles, in P. Mandl and M. Huskova (eds.), Asymptotic Statistics, 349--359, Springer-Verlag, New York.
[2] Kocherginsky, M., He, X. and Mu, Y. (2005). Practical Confidence Intervals for Regression Quantiles, Journal of Computational and Graphical Statistics, 14, 41-55.
[3] Hagemann, A. (2017) Cluster Robust Bootstrap inference in quantile regression models, Journal of the American Statistical Association , 112, 446--456.
[4] He, X. and Hu, F. (2002). Markov Chain Marginal Bootstrap. Journal of the American Statistical Association , Vol. 97, no. 459, 783-795.
[5] Parzen, M. I., L. Wei, and Z. Ying (1994): A resampling method based on pivotal estimating functions,'' Biometrika, 81, 341--350.
[6] Bose, A. and S. Chatterjee, (2003) Generalized bootstrap for estimators of minimizers of convex functions, J. Stat. Planning and Inf, 117, 225-239.
[7] Chamberlain G. and Imbens G.W. (2003) Nonparametric Applications of Bayesian Inference, Journal of Business & Economic Statistics, 21, pp. 12-18.
[8] Feng, Xingdong, Xuming He, and Jianhua Hu (2011) Wild Bootstrap for Quantile Regression, Biometrika, 98, 995--999.
summary.rq
y <- rnorm(50)
x <- matrix(rnorm(100),50)
fit <- rq(y~x,tau = .4)
summary(fit,se = "boot", bsmethod= "xy")
summary(fit,se = "boot", bsmethod= "pwy")
#summary(fit,se = "boot", bsmethod= "mcmb")
Run the code above in your browser using DataLab