R/sequence_thresh.R

In Sorbus-torminalis/twopointzerothree: Detects Duplicate Sequences of Offset Values

#' Threshold segment correlations
#' @description Thresholds correlation between all segments length n of a vector
#' @param sequ Output of [twopointzerothree::sequence_cor()]
#' @param threshold Minimum correlation to accept. Defaults to 1, possible range -1 to 1 (but values much below 1 make no sense)
#' @param tolerance Small positive number to allow for numerical imprecision
#' @details Tests if correlations are >= than a given threshold (with tolerance for imprecision in correlations). Default threshold is 1 (perfect correlation). Might be useful to reduce this slightly, or equivalently increase tolerance to allow for cases where imprecision is high.  
#' @examples
#' set.seed(42) 
#' x <- rnorm(10)
#' x <- c(x, x[1:4] + 2.03)
#' sequ <- sequence_cor(vec = x, n = 4)
#' sequence_thresh(sequ)    
#' @export

sequence_thresh <- function(sequ, 
                           threshold = 1, 
                           tolerance = sqrt(.Machine$double.eps)){ 
  if(tolerance < 0){
    stop("tolerance must be positive")
  }
  if(threshold < -1 | threshold > 1){
    stop("thresh must be between -1 and 1 
         (only values close to 1 make sense)")
  }
  th <- sequ + tolerance >= threshold
  attr(th, "len") <- attr(sequ, "len")
  th
}