Statistical aspects of aquaculture research: sample sizes for pond experiments

The present paper illustrates the need for greater pond replication in pond experiments. It demonstrates that increasing the number of fish sampled in each pond will not reduce the number of ponds needed when treatment effects are estimated. Sampling bias, skew distributions and heterogeneous fish populations are discussed in relation to fish sample sizes. When it is impossible to increase the number of replicate ponds, several methods of accounting for the pond variability are suggested.