Eliminating pond differences with cross-over designs

Large variability between ponds in aquaculture experiments often leads to imprecise estimates of treatment effects. One solution would be to increase the number of experimental ponds but the availability of ponds is usually limited. A cross-over trial makes comparisons directly within the same pond, instead of across ponds, thus taking account of individual pond effects and potentially removing much of the variability. A cross-over trial was initiated to test the feasibility of using such designs in pond aquaculture research. When compared to a completely randomized design, the cross-over trial was more efficient for comparing the fish growth for each of the two treatments, and, on average, a considerable reduction in the standard error of the treatment estimate was achieved. Thus, under certain circumstances, the cross-over design has great potential to improve the precision of pond aquaculture experimentation.