45 lines
1.4 KiB
R
45 lines
1.4 KiB
R
#' Conditional Variance Estimator
|
|
#'
|
|
#' Conditional Variance Estimator (CVE) is a novel sufficient dimension
|
|
#' reduction (SDR) method for regressions satisfying E(Y|X) = E(Y|B'X),
|
|
#' where B'X is a lower dimensional projection of the predictors.
|
|
#'
|
|
#' @param X A matrix of type numeric of dimensions N times dim where N is the number
|
|
#' of entries with dim as data dimension.
|
|
#' @param Y A vector of type numeric of length N (coresponds to \code{x}).
|
|
#' @param k Guess for rank(B).
|
|
#' @param nObs As describet in the paper.
|
|
#'
|
|
#' @param tol Tolerance for optimization stopping creteria.
|
|
#' @export
|
|
#'
|
|
#' @seealso TODO: \code{vignette("CVE_paper", package="CVE")}.
|
|
#'
|
|
#' @references Fertl Likas, Bura Efstathia. Conditional Variance Estimation for Sufficient Dimension Reduction, 2019
|
|
cve <- function(X, Y, k,
|
|
nObs = sqrt(nrow(X)),
|
|
tauInitial = 1.0,
|
|
tol = 1e-3,
|
|
slack = -1e-10,
|
|
maxIter = 50L,
|
|
attempts = 10L
|
|
) {
|
|
# check data parameter types
|
|
stopifnot(
|
|
is.matrix(X),
|
|
is.vector(Y),
|
|
typeof(X) == 'double',
|
|
typeof(Y) == 'double'
|
|
)
|
|
|
|
# call CVE C++ implementation
|
|
return(cve_cpp(X, Y, k,
|
|
nObs,
|
|
tauInitial,
|
|
tol,
|
|
slack,
|
|
maxIter,
|
|
attempts
|
|
))
|
|
}
|