Evaluation of Candidate Net Selectivity Models for 1990-2003 Yukon River Sonar Gill-net Catch Data. Alaska Fisheries Technical Report No. 75

摘要

Thirty eight net selectivity models were evaluated for their utility in representing catch data of a variety of fish species at the Yukon River sonar project on the mainstem Yukon River near Pilot Station, Alaska. Gill-net catch data collected from 1990 to 2003 were divided into 9 data sets representing species groups and each of the 38 models was fit to each of the 9 data sets. Model parameters were estimated using maximum likelihood techniques. Model utility was evaluated by comparing maximized values of the loglikelihood function, Akaikes Information Criterion statistics, scaled deviance statistics, and two types of diagnostic plots. One family of models almost universally provided the best representation of the catch data for all 9 species groups. One model from this family of models is recommended for use to apportion Yukon River sonar estimates of total fish abundance to species.