SAMPLING FROM TRAWL GEARS USED IN SIZE-SELECTIVITY EXPERIMENTS

When the catches in the components (cod-end and cover, say) of experimental trawl selectivity gear are sampled then the length-class frequencies of the catches can be estimated by multiplying the frequencies in the samples by the inverse of the sampling fractions. The statistical properties of selection curves fitted to these estimated catch frequencies are unknown, and consequently goodness of fit cannot be checked and trawl selectivity studies cannot be designed. It is shown that a statistically well-behaved procedure is obtained by working directly with the raw data whereby the estimated selection curve is obtained from the fitting of a related curve to these raw data. For covered cod-end trials in which a logistic selection curve is assumed the related curve is also logistic and no modifications to existing software are required. This facilitated an investigation of the efficiency of three different sampling strategies: to sample from same number of fish; or (3) the same fraction of fish. Strategy (1) was found to be the least efficient. Strategies (2) and (3) had similar performance, with strategy (2) slightly better overall.