Effective experimental designs for optimizing fish feeds

Determination of the optimal protein-to-energy ratio of fish feed is used as an example of how to improve the efficiency of experimental designs for feed formulation research. Studying one variable at a time is inefficient and lacks information on interactions between feed components. Even sound designs, such as full factorials, may fail in feed formulation studies because often two or more factors are implicitly confounded. A class of response surface methodology, namely designs for mixture experiments, is well suited for feed formulation problems and ensures that results are readily applicable to the feed industry. Combined with some criteria for searching the optimal runs for the experiment, mixture designs can improve the efficacy of resource use and diminish risks of inference and decision-making in fish feed research and development.