R/remove_consecutive_jumps.R

In YalDan/hf.econometrics: High Frequency Econometrics

#' Remove consecutive jumps
#'
#' Helper function to remove consecutive jumps
#'
#' @import data.table
#'
#' @param DATA The output from `LM_JumpTest_2012`.
#' @param RANGE_OBS How many observations away must a jump be to be considered non-consecutive? Defaults to 3, s.t. all jumps in the next 3 blocks will be removed.
#' @param split_vars By which variables should the data be split? Defaults to `c("date", "hour", "s")`.
#'
#' @return Returns the data without consecutive jumps

#' @export
remove_consecutive_jumps <- function(DATA, RANGE_OBS = 3, split_vars = c("date", "hour", "s")){

  #### load data and split it ####
  split_jumps <- split(DATA[Jump_indicator != 0], by = split_vars)
  ###

  ###
  split_no_consec_jumps <- lapply(seq_along(split_jumps), function(x)
  {
    ## take subset ##
    DT_tmp <- split_jumps[[x]]

    ## reverse dataset to iterate backwards ##
    DT_tmp_rev <- DT_tmp[seq(dim(DT_tmp)[1],1),]

    ## preassign iterator ##
    if (nrow(DT_tmp_rev) > 1) {
      n_iter <- 1:(nrow(DT_tmp_rev) - 1)
    } else {
      n_iter <- 1
    }

    ## loop over all rows ##
    for (i in n_iter) {
      if(!is.na(DT_tmp_rev[i+1,count])){ # only iterate if last row is not reached

        #  remove all rows that contain a consecutive jump #
        # a consecutive jump is measured if jump is in the neighborhood of RANGE_OBS blocks
        if(DT_tmp_rev[i, count]-DT_tmp_rev[i+1, count] <= RANGE_OBS*(DT_tmp_rev[1,kM])){
          DT_tmp_rev <- DT_tmp_rev[!i]
        }
      }
    }

    ## undo reversal ##
    DT_tmp <- DT_tmp_rev[seq(dim(DT_tmp_rev)[1],1),]
  })
  return(rbindlist(split_no_consec_jumps))
}