Marine monitoring programs provide data that are essential for marine management. The reliability of such data is underpinned by proficiency tests. In the context of Quasimeme, a proficiency testing program for the marine environment, a statistical model has been developed in 2000 to evaluate data sets. The mathematical basis of this model has been reformulated in 2016. In this article, the new mathematical basis is concisely explained. The main results of a comparative study using five statistical procedures on about 2400 data sets are described. Examples of the application of the five techniques on two data sets with complicated distributions are given. It is concluded that the NDA implementation of our model offers the highest robustness and that a thorough evaluation should invoke graphical representations of the data sets and consider robust skewness and results from different statistical methods.