where cp is a multiplier to give the 95th percentile of a normal distribution (corrected for use of the biased observed SD estimate instead of the true but unknown population SD), B = total number of blank results in the dataset, and K = number of blank samples.