How many pigs? Statistical power considerations in swine nutrition experiments

ABSTRACT Replication refers to the assignment of more than one experimental unit to the same treatment. Each replication of a treatment is an independent observation; thus, each replication involves a different experimental unit. In swine nutrition research, the experimental unit may be an individual animal, as in sow reproduction experiments, or a group of animals, as in growing-finishing pig experiments. In either case, calculation of the number of replicates needed to give an accurate and reliable outcome is an important step in deriving an experimental protocol. Although investigators often seem to choose replication arbitrarily on the basis of cost or availability of animals, housing considerations, convenience, or tradition, the question of how many pigs (i.e., how much replication is necessary) is a statistical one that has a statistical answer. A power analysis, performed while designing an experiment, will provide an investigator with an estimate of the number of replicates needed for an experiment of known power and sensitivity. This a priori, or prospective, power analysis ensures that an investigator does not waste time and resources carrying out an experiment that has little chance of finding a significant effect, if one exists. It also makes sure resources are not wasted by including more experimental units than are necessary to detect an effect. A retrospective, or a posteriori, power analysis may also be conducted. If no significant effects are found in an experiment, an investigator can assess the observed power of the experiment, or may determine the size of treatment effect that could have been detected using the standard deviation and number of replicates in the experiment. The latter may be useful in explaining results. However, the former may be misleading because a high P-value will invariably result in a low observed power, and little new information will be gained from the post hoc power analysis. In most cases, the time for making power calculations is before, not after, an experiment is conducted.

Key Words: Replication, Statistical Power, Swine Nutrition

© 2004, by the American Society of Animal Science. All rights reserved.

J. Anim. Sci. 2004. 82(E. Suppl.):E245-E254

Implications

Careful planning and organization before an experiment is begun can maximize the amount of information gained. Investigators should focus on using an a priori, or prospective, power analysis to determine the appropriate number of replications, during the planning stage of an experiment. When nonsignificant results are found during the analysis stage, an a posteriori, or retrospective, power analysis can be useful for determining the size of the difference that could have been detected given the size of the experiment and the observed variability. Calculating observed power after an experiment has been conducted and nonsignificant results obtained contribute no new information, and may, in fact, be misleading.