A recent article (here – open access) gives an overview of non-targeted approaches to authenticity screening. It is written at a level of technical detail useful to laboratory analysts or technically-informed industry customers. It discusses spectroscopic techniques (NMR, IR and MS) and the use of machine learning to derive classification sets from reference sample sets. It focusses, particularly, on how to deal with the statistical probability of a subsequent false-accept or a false-reject verdict on a test sample based upon an incorrect or uncertain (“grey”) classification.
Photo by Markus Winkler on Unsplash