R/ease_swamping.R

In hidetify: Identify Influential Observations in High Dimension

ease_swamping <- function(
                            x,
                            y,
                            xquant, 
                            yquant, 
                            inv_rob_sdx, 
                            rob_sdy,number_subset,
                            size_subset,
                            est_clean_set,
                            asymvec,
                            ep=0.1,
                            alpha
                            )
{
  n <- nrow(x)
  min_sum_Him <- rcpp_mask_swamp_stat(x, y, xquant, yquant, inv_rob_sdx, 
                                      rob_sdy, number_subset, size_subset, 
                                      est_clean_set, asymvec) # rcpp function
  
  min_sum_Him <- c(min_sum_Him[,1])
  pvv <- pchisq(min_sum_Him, df=1, lower.tail=FALSE)
  Spvv <- sort.int(pvv,index.return=TRUE)
  Sii <- Spvv$ix
  dpv <- Spvv$x-alpha/length(min_sum_Him)

  In <- which(dpv <= 0)
  if (length(In) == 0)
  {clean_set <- est_clean_set}
  else
  {
    rin <- max(In)
    inf_setv <- est_clean_set[Sii[1:min(floor(ep*n),rin)]]
    clean_set <- setdiff(est_clean_set,inf_setv)
  }

  est_clean_set <- clean_set
  list(est_clean_set = est_clean_set)
}