Repeated measures experiments in forestry: focus on analysis of response curves

Abstract

Treatment effects over time are frequently investigated using repeated measures designs, but analyses of these experiments frequently fail to address a primary objective of collecting data over time, namely description of the response curve. The analysis advocated in this paper utilizes the intrinsic continuity of the repeated measures factor by focusing on response curves. Treatments are compared by analyzing estimated coefficients of response curves proposed by the investigator. This approach provides more information on treatment effects than analyses that compare treatments separately at each time period. Analysis of estimated coefficients is easier to interpret than multivariate analyses of variance and does not require often biologically implausible assumptions of split-plot analyses currently in vogue. An example describing effects of aluminum on sugar maple (Acersaccharum Marsh.) seedling growth illustrates the method.