2019-09-16 09:15:51 +00:00
|
|
|
% Generated by roxygen2: do not edit by hand
|
|
|
|
% Please edit documentation in R/estimateBandwidth.R
|
|
|
|
\name{estimate.bandwidth}
|
|
|
|
\alias{estimate.bandwidth}
|
2019-10-18 07:06:36 +00:00
|
|
|
\title{Bandwidth estimation for CVE.}
|
2019-09-16 09:15:51 +00:00
|
|
|
\usage{
|
|
|
|
estimate.bandwidth(X, k, nObs)
|
|
|
|
}
|
|
|
|
\arguments{
|
2019-10-18 07:06:36 +00:00
|
|
|
\item{X}{data matrix with samples in its rows.}
|
2019-09-16 09:15:51 +00:00
|
|
|
|
2019-10-18 07:06:36 +00:00
|
|
|
\item{k}{Dimension of lower dimensional projection.}
|
2019-09-16 09:15:51 +00:00
|
|
|
|
2019-10-18 07:06:36 +00:00
|
|
|
\item{nObs}{Expected number of points in a slice, see paper.}
|
2019-09-16 09:15:51 +00:00
|
|
|
}
|
|
|
|
\description{
|
|
|
|
Estimates a propper bandwidth \code{h} according
|
|
|
|
\deqn{%
|
2019-10-18 07:06:36 +00:00
|
|
|
h = \chi_{k}^{-1}\left(\frac{nObs - 1}{n-1}\right)\frac{2 tr(\Sigma)}{p}}{%
|
|
|
|
h = qchisq( (nObs - 1)/(n - 1), k ) * (2 tr(\Sigma) / p)}
|
|
|
|
with \eqn{n} the number of sample and \eqn{p} its dimension
|
|
|
|
(\code{n <- nrow(X); p <- ncol(X)}) and the covariance-matrix \eqn{\Sigma}
|
|
|
|
which is given by the standard maximum likelihood estimate.
|
2019-09-16 09:15:51 +00:00
|
|
|
}
|
|
|
|
\seealso{
|
|
|
|
[\code{\link{qchisq}}]
|
|
|
|
}
|