Statistical power (and conversely, Type II error) is often ignored by biologists. Power is important to consider in the design of studies, to ensure that sufficient resources are allocated to address a hypothesis under examination. Deter- mining appropriate sample size when designing experiments or calculating power for a statistical test requires an investigator to consider the importance of making incorrect conclusions about the experimental hypothesis and the biological importance of the alternative hypothesis (or the biological effect size researchers are attempting to measure). Poorly designed studies frequently provide results that are at best equivocal, and do little to advance science or assist in decision making. Completed studies that fail to reject Ho should consider power and the related probability of a Type II error in the interpretation of results, particularly when implicit or explicit acceptance of Ho is used to support a biological hypothesis or management decision. Investigators must consider the biological question they wish to answer (Tacha et al. 1982) and assess power on the basis of biologically significant differences (Taylor and Gerrodette 1993). Power calculations are somewhat subjective, because the author must specify either f or the minimum difference that is biologically important. Biologists may have different ideas about what values are appropriate. While determining biological significance is of central importance in power analysis, it is also an issue of importance in wildlife science. Procedures, references, and computer software to compute power are accessible; therefore, authors should consider power. We welcome comments or suggestions on this subject.
|Authors||M.J. Conroy, M. D. Samuel, Joanne C. White|
|Publication Subtype||Journal Article|
|Series Title||Journal of Wildlife Management|
|Record Source||USGS Publications Warehouse|
|USGS Organization||National Wildlife Health Center|