Simulation-based evaluation of the accuracy in indirectly estimating gillnet selectivity

We used stochastic simulations to evaluate the accuracy in indirectly estimating selectivity for experimental gillnets and to gage how estimator performance was influenced by factors such as sample size and gillnet configuration. Included in our evaluations were cases where both selection curve model parameters and relative fishing intensities of gillnet panels were estimated simultaneously. We additionally explored how estimation accuracy was affected by assuming incorrect relative fishing intensities when fitting selectivity models. For binormal and lognormal selection curve models, parameter estimates were generally accurate and precise when intensities were fixed at their true values, although some biases could result in estimates of model spread under small sample sizes. Conversely, when relative fishing intensities were estimated simultaneously with binormal selection curve model parameters, substantial bias and imprecision could result for both selection curve parameters and intensities depending on sample sizes. Assuming equal relative fishing intensities for gillnet panels when intensities actually varied resulted in some estimator bias of selection curve model parameters depending on the underlying intensity pattern and assumed age composition and growth patterns of the population. In the case of binormal selection curve models, overall performance of parameter estimates was often as good as or better than when relative fishing intensities were estimated. Given the potential problems in estimating relative fishing intensities in gillnet selectivity evaluations, including the potential of parameter confounding for some selection curve models and high bias and imprecision with small sample sizes for other selection curve models, we do not recommend this as a standard approach when indirectly estimating gillnet selectivity.