Location: Dairy Forage ResearchTitle: Optimizing Experimental Designs: Finding Hidden Treasure.) Author
Submitted to: ASA-CSSA-SSSA Annual Meeting Abstracts
Publication Type: Abstract only
Publication Acceptance Date: 5/31/2008
Publication Date: 10/5/2008
Citation: Casler, M.D. 2008. Optimizing Experimental Designs: Finding Hidden Treasure [abstract]. ASA-CSSA-SSSA Annual Meeting Abstracts. Abstract No. 722-2. Interpretive Summary:
Technical Abstract: Classical experimental design theory, the predominant treatment in most textbooks, promotes the use of blocking designs for control of spatial variability in field studies and other situations in which there is significant variation among heterogeneity among experimental units. Many blocking designs fail to remove significant amounts of spatial variability, because it is impossible, in most situations, to accurately predict the optimal size, shape, and orientation of spatial blocks. In effect, this result reduces precision on treatment mean comparisons, increasing the size of differences that can be detected with a desired P-value. Hidden treasure refers to useful information on spatial variability that, with a little bit of training and experience, can be extracted from historical field trials. The more historical field trials that have been conducted at a particular site or, better yet, in a particular field, the more useful this information will be for optimizing experimental designs in future trials. First, postdictive assessments of blocking designs can provide valuable information to determine (1) the effectiveness of the blocking design per se, (2) whether or not changes in plot size are expected to be effective, and (3) the relative importance of multiple precision-enhancing strategies (increasing the number of reps, increasing plot size, decreasing block size). Second, multiple statistical approaches are available for postdictive analysis of spatial variation on a plot-by-plot basis, regardless of the blocking patterns employed in the existing design. Spatial analyses, including nearest neighbor analyses, trend analysis, and power analyses, can be used to adjust treatment means and standard errors for spatial variability and to inform future decisions on optimal blocking designs for a particular field site.