Evaluation of censored contamination data†

Abstract

Laboratory measures of low‐level contamination are typically strongly skewed, and observations below a specified detection limit are recorded merely as not detectable or less than a specified limit. In these circumstances, a full description of the distribution requires assumptions about the form of the lower tail. If such assumptions are made, the distribution can be described in terms of the usual parameter estimates. The assumptions can be checked if similar samples can be accurately assessed using specialized equipment with a very low detection limit. Methods of fitting suitable distribution to data of this sort are discussed, and evaluated using results from a laboratory with equipment capable of measuring very low contamination levels. Estimates based on the complete data are compared with estimates based on assumed detection limits.