A reply to: Some comments on FISHERIES/2020/JAN/SWG-PEL/08
Document FISHERIES/2020/JAN/SWG-PEL/08 provided suggestive indications related to two aspects on estimates of the impacts on penguins of pelagic fishing in the neighbourhood of islands containing penguin breeding colonies based on models using individual penguin response data for input. These were that the associated variance estimates for models that had been presented by that date were not robust to the choice made for the random effects structure, and further that the structure of the data themselves was such that use of individual data would lead to very little if any gain in precision for estimates of closure effects compared to estimates based on annually aggregated data. The comments in FISHERIES/2020/AUG/SWG-PEL/83 (Sherley) fail to address either of these concerns. First, Sherley’s own results in Table 1 of FISHERIES/2020/JUL/SWG PEL/53REV serve to confirm that concerns that variance estimates would lack robustness to the choice of the random effects structure were justified, and hence that earlier estimates presented by Sherley and co-authors which failed to address the matter of an appropriate choice for this structure are not reliable. Secondly, the indication that multiple within-year data provide no additional information on the between-year island-closure effect is not addressed. Document FISHERIES/2020/AUG/SWG-PEL/83 therefore provides no evidence to invalidate the conclusion of FISHERIES/2020/AUG/SWG-PEL/82, based especially on the more detailed mathematical-statistical analyses of its Annex, that estimation based on individual data is unable in principle to provide any improvement in precision of the results from analyses of annually aggregated data. While the management implications of results from the island closure experiments do merit consideration, unless a mathematical demonstration can be provided that the analyses of FISHERIES/2020/AUG/SWG-PEL/82 are incorrect, such consideration must disregard results based on the estimation approach using individual data put forward by Sherley and co-authors.
