University of Cape Town
MARAM_IWS_DEC14_PENG_B15.pdf (196.3 kB)

Taking Account of Sample Size in the GLM Analyses of the Response Variable Data from the Island Closure Feasibility Study

Download (196.3 kB)
posted on 2022-03-25, 13:17 authored by Doug ButterworthDoug Butterworth, Andrea Ross-GillespieAndrea Ross-Gillespie
Two approaches are used to examine the impact of taking sample size into account in GLM analyses to estimate the fishing effect parameter λ. In the first the variance of each mean value input to the analyses is disaggregated into separate contributions reflecting process error (taken to be constant) and observation error (taken to be inversely proportional to sample size). The second simpler approach merely omits values for which the sample size was very low. The implications are evaluated for the majority of the scenarios considered and methods applied in earlier analyses which did not account explicitly for sample size. The pattern of results when sample size is taken into account is clear and consistent: broadly speaking results do not change much in the great majority of cases, and in particular the substantial preponderance of positive to negative estimates of the fishing effect parameter λ remains



Department of Mathematics and Applied Mathematics, University of Cape Town