Repeated Measures Analyses for Tree Fruit Experiments
Experiments with perennial crops often span several years, and a response variable may be measured on the same plant at several points in time. Such data are often analyzed as a split-plot design, taking time as the split-plot factor. In other cases, separate analyses are performed for each time. The mathematical conditions required for validity of these types of analyses might not hold because measurements repeated on the same plant are not independent. Annual trunk cross-sectional-area (TCSA) measurements from a peach tree training experiment will be used to compare two methods of analyses. The 6-year experiment was a factorial of two heading heights at planting (low vs. high) and two tree forms (central leader vs. open vase). Univariate analysis of variance (ANOVA) and a multivariate repeated measures analysis (MANOVA) was performed. Main effects and interactions were more often significant with ANOVA than with MANOVA. ANOVA performed each year inflated the probability of falsely rejecting a true null hypothesis (Type I error), and was not appropriate for this data set.