SREL Reprint #1774
Statistical issues for field experimenters
Philip M. Dixon and Karen A. Garrett
Abstract: This chapter reviews four frequently misunderstood statistical aspects of the design and analysis of field experiments: generality of conclusions, pseudoreplication, statistical power, and testing for no effect. The choice of design determines the appropriate inference space (degree of generality of the conclusions) and the appropriate error terms to test hypotheses about treatments. Pseudoreplication, one form of use of inappropriate error terms, is common in analyses of field experiments, but it can be avoided by applying replicate treatments or restricting the generality of one's conclusions.
Statistical power is the ability of an experiment to detect differences between treatments. Field experiments frequently have low power because of few replicates and large residual variation. Power may be increased by changing the design, reducing the residual variation, or increasing the number of replicates.
Finally, bioequivalence tests are discussed, which are appropriate to test that a particular treatment has no effect. This goal is the reverse of the goal of a traditional hypothesis test, which is to show that a treatment has an effect. Because bioequivalence tests are not widely used, the authors demonstrate their use and interpretation with an example.
Keywords: statistical analysis, experimental design, bioequivalence tests, power, sample size, pseudoreplication
SREL Reprint #1774
Dixon, P. M. and K. A. Garrett. 1994. Statistical issues for field experimenters. pp. 439-450 In: R. J. Kendall and T. E. Lacher, Jr. (Eds.). Wildlife Ecology and Population Modeling: Integrated Studies of Agroecosystems. Lewis Publishers, Boca Raton, FL.
This information was provided by the University of Georgia's Savannah River Ecology Laboratory (srel.uga.edu).