An evaluation of survey design for estimating the food consumption by fish

Abstract

The effect of survey design on the precision of estimates of average fish stomach contents is examined. The contribution to the total variance from within and between tow variability in stomach contents is evaluated, and the implications for stomach sampling programs are discussed. As an example we have estimated the average amount of capelin in Barents Sea cod stomachs for years of very low, low and medium capelin abundance. The results indicate that to maximize precision for a given cost it is generally best to sample stomachs at as many locations as possible. A simulation study based on resampling from these data suggests that little is gained in precision by collecting 5 instead of 2 stomachs from each 5 cm length group of fish.