Practical analysis of factorial experiments in forestry

Abstract

Factorial designs are among the most frequently employed for arranging treatments in forestry experiments. Yet researchers often fail either to recognize the factorial treatment structure or to take full advantage of the structure for interpreting treatment effects. The analysis of factorial experiments should focus on comparisons of means of research interest specified by the investigator. Reliance on default options of computing packages or routine application of multiple comparison procedures often fails to address research hypotheses directly. A two-step strategy for the analysis of factorial experiments entails a check for interaction followed by estimation of either main effects or simple effects. This strategy emphasizes sensible mean comparisons through estimation of contrasts and their standard errors. The strategy also applies to the analysis of factorial experiments in which unequal replication or empty cells complicate the analysis. We summarize a practical approach for use by forest scientists and applied statisticians consulting with such scientists so that they may analyze and interpret their experiments more effectively.