An initial implementation of suggestions by the 2020 Panel to improve estimates of the effect of fishing around islands on penguins by using models with random effects and applied to both aggregated and disaggregated data
reportposted on 03.05.2021, 15:18 by Andrea Ross-GillespieAndrea Ross-Gillespie, Doug ButterworthDoug Butterworth
Pending availability of data on the month for each observation for further analyses of the island closure experiment to address the suggestions made by the 2020 Panel, some initial analyses are conducted to explore the implications of the Panel’s suggestions regarding sample size weighting for data-aggregated approaches, and nesting of random effects for methods using disaggregated data. These are applied to existing data for maximum foraging distance and chick condition. Results indicate that when these Panel suggestions are taken into account, there is little to no real difference in the results which these two approaches provide for the effect of fishing parameter. This is as the Panel anticipated, and in line with an earlier mathematical demonstration that use of the extra data available under the disaggregated approach would not improve the precision of these estimates. Some corresponding earlier results suggesting greater precision for the disaggregated approach were a consequence of applying a random effects approach without appropriate nesting of the data – an approach which is rejected by statistical analyses. Results for estimates of the effect of fishing parameter are more meaningfully expressed in terms of the consequent impact on the penguin population growth rate. For the cases examined initially here, the only meaningful result suggesting a negative impact of fishing in the vicinity of islands is for the maximum foraging distance variable for St Croix island.