In 2009, a self-sampling programme was organized in the Netherlands, fishers sampling ca. 80 kg of discards from randomly selected bottom trawls in the North Sea. A statistical procedure is proposed to highlight samples, trips (with multiple samples), or vessels (which may have multiple trips within a year) where extreme mean lengths of discarded fish were observed. Randomization methods were used to test for evidence of non-randomness in patterns of highlighted discard samples, e.g. repeated observations of extreme mean lengths for consecutive discard samples across trips from the same vessel. European plaice (Pleuronectes platessa), common dab (Limanda limanda), grey gurnard (Eutrigla gurnardus), and whiting (Merlangius merlangus) were considered because these were the most abundant species in most of the discard samples. A linear mixed model was used to estimate randomsample effects on the estimated mean lengths by species. These random effects were incorporated into uni- and bivariate procedures to identify extreme samples that were summed for each vessel, and the probability of observing such numbers was estimated. Excluding these samples from the dataset had marginal effects on estimated size distributions of fish.