Conclusions The chemical interpretation of the detection limit and its stochastic model counterpart has thus different consequences for the statistical analysis than we would expect from the word by word interpretation of “below detection limit” as a concentration below some limit. The state of the art model on BDL compositional analysis is biased. Some ideas are not directly applicable to the true effects of measurement errors near the detection limit. Even the basic principles like subcompositional coherence and the requirement of the independence of the analysis from the total are not fully valid near the detection limit.

Peer Reviewed