R/ROT.R

In ADPclust: Fast Clustering Using Adaptive Density Peak Detection

##' Calculate the ROT bandwidth either from a data frame, or from p and n. 
##' 
##' IMPORTANT NOTE: The standard deviation of each variable is omitted in this formula.
##'
##' @title Calculate ROT bandwidth
##' 
##' @export
##' 
##' @param x the number of variables (if y is missing), or a data frame or a matrix (if y is not missing).
##' @param y the number of observations. If y is missing, x should be the data matrix.
##' @return ROT bandwidth.
ROT <- function(x, y = NULL){
    if(is.null(y)){
        if(inherits(x, c("data.frame", "matrix"))){
            n <- nrow(x)
            p <- ncol(x)
        }else{
          stop("y is missing. x is must be a data frame.")
        }
    }else{
        if(is.numeric(x) && x > 0 && is.numeric(y) && y > 0){
            p = x
            n = y
        }else{
            stop("Wrong x, y.")
        }
    }
    h <- n ^ (-1 / (p + 4)) ## Scott's ROT
    return(h)
}