A PARAMETRIC MULTINOMIAL MODEL FOR SIZE SELECTION IN ALTERNATE-HAUL EXPERIMENTS

A parametric model for estimating mesh size selectivity is derived fro m the maximum likelihood estimation method of Skalski and Perez-Comas (1993). It is based on a multinomial likelihood for sampling from a mi xture of discrete size classes, and assumes logistic curves to describ e relative size selectivity. Analysis was illustrated for haddock (Mel anogrammus aeglefinus), pollock (Theragra chalcogramma) and petrale so le (Eopsetta jordani) caught in alternate-haul experiments. Results we re contrasted with those from three other methods (Pope et al., 1975; Simpson, 1989; Millar, 1992). For the three examples, the performance of the multinomial method was statistically acceptable and similar to that of the Millar method. Possible sources of bias due to departures from the model assumptions are discussed.