The statistical properties of a stepwise procedure for testing the purity of fruit juice

In many cases no fixed standards are available and a sample from a target population is used to provide the necessary quality-control norms. Furthermore, many products possess several correlated attributes that are relevant to the authenticity question. This article addresses some issues related to statistical tests used in such cases, focussing in particular on a procedure that has been lately proposed for differentiating between pure and adulterated fruit juice. The quality control of fruit juice is usually performed by comparing the values of several characteristics in the tested sample against some prespecified standards. In many cases those standards are based on the values found in a sample of a presumably pure juice, which is used as a "reference sample" or a "base sample". Several detection methods were recently proposed that use universal "base samples" and are supposed to be valid when applied to juices from different sources or different varieties. We compare the results yielded by the use of various "base samples" and show that the application of such a method to Israeli citrus juice leads to too many rejections of pure juices. We suggest that the main drawbacks of the method is the use of improper "base sample" that is not specific to the sample being tested.