Sample size recommendations for estimating stock composition using genetic stock identification (GSI)

"Genetic Stock Identification (GSI) allows estimation of stock composition, including untagged stocks. The required sample size per sampling stratum depends on the commonness/rarity of the stock(s) of interest and, when the stock composition of the landings itself is of interest rather than the ocean mixed stock aggregate, on total catch as well. We present sample sizes required to achieve specified precisions, detect rare stocks, and test the significance of differences between proportions. For the ocean mixed stock aggregate, sampling 800 fish per stratum suffices to estimate stock proportions 3% with a coefficient of variation 20% and has a high (> 99 : 9%) probability of detecting stocks comprising 1% of the mixed stock, but this sample size does not yield precise estimates of small stock proportions and may fail to detect rare stocks. To precisely estimate smaller proportions, required sample size scales approximately inversely with the proportion, i.e., halving the target proportion requires doubling the sample size. For a given proportion, increased precision also comes at a greater cost as the required sample size scales approximately with the square of the desired precision, i.e., halving the target standard error requires quadrupling the sample size. Smaller sample sizes suffice for estimating the stock composition of the landings itself when total catch is low, but a high sampling fraction may be required. Statistical comparisons between two proportions require larger sample sizes than estimating single proportions of similar magnitude"--Abstract. Shanae D. Allen-Moran, William H. Satterthwaite, Michael S. Mohr. "June 2013." System requirements: Adobe Acrobat Reader. Includes bibliographical references (pages 20-22).