Peter Ward and Ransom A. Myers. 2005. Shifts in open-ocean fish communities coinciding with the commencement of commercial fishing. Ecology 86:835-847.

Appendix B. Parameter estimates and diagnostic statistics for generalized linear models used to estimate relative abundance.

Introduction

This appendix provides details of the generalized linear models that we used to develop the indices of abundance presented in our Ecology article. We also present estimates and statistics for models that were used to investigate the robustness of abundance indices. Appendix A describes those investigations and the data that were fitted. This Appendix includes output from models that were not presented in the Ecology article because they did not meet the criteria of having the lowest standard error that was less than one with an estimate of the fishing period coefficient that was significantly different to one. The model output was generated by SAS (version 8.0) using the GENMOD procedure.

Unlike the coefficients presented in our Ecology article and Appendix A, the estimates of the coefficient of fishing period in this appendix were not adjusted for differences in sampling between periods. Table B1 lists the adjustments that should be applied to the fishing period coefficient of each model. Adjustments were based on the ratio R of the numbers caught and modeled in each period:

                                                       (B.1)

where T is the total number of the species caught in the study region and M is the number modeled. For the hook model, for example, the ratio of the numbers of albacore tuna caught and modeled was 0.811 and the estimate of the fishery period coefficient was –2.730. Therefore the change in abundance ΔA of albacore tuna was estimated to be:

                                                    (B.2)

We applied a ratio of one to estimates where no species were caught in a period, e.g., pelagic stingray (Dasyatis violacea) and pomfrets (Bramidae).

TABLE B1. The numbers of each species caught and modeled. We used the ratio of the numbers caught and modeled in each period (Eq. B.2) to adjust the estimate of each fishery coefficient, which are presented in our Ecology article and Appendix A.