A new statistic for sampling design investigation : an application to length-structured landings sampling

No abstracts are to be cited without prior reference to the author.

Measuring fish on a market or on board a vessel is the most common action taken by biologists all around the world to gather data for stock assessment. The samples coming from this kind of survey are of vector type and are not straightforward to analyse and compare. This paper proposes a generic tool that enables to investigate in any multinomial like sampling scheme. The use of this tool will be discussed based on examples from two different sampling designs. One is a sampling of commercial market categories and the other is a métier based sampling. The research of outliers, misallocated samples or potential bias, the stratification and the analysis of within and between strata variance will be discussed with the objective of constructing the best sampling design and insure the quality of the input data for stock assessment models. This statistic is easy to implement and should be considered as a mandatory step to validate the sampling data at an international level.