Submitted to: Communications in Biometry and Crop Science (CBCS)
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: August 13, 2013
Publication Date: September 9, 2013
Citation: Casler, M.D. 2013. Finding hidden treasure: a 28-year case study for optimizing experimental designs. Communications in Biometry and Crop Science (CBCS). 8 (1), 23–38. Interpretive Summary: Agronomic and breeding research requires the extensive use of field plots. Due to non-uniformity of agricultural fields at research stations, researchers are required to use various methods of organizing experiments and collecting data that create uniform conditions for comparing treatments, systems, or genotypes. This paper summarizes 28 years of forage grass field trials at Arlington, Wisconsin. Manipulations employed to create uniformity during this time period include: changes in the way that plots are randomized, changes in block size, changes in plot size, and changes in the methods of statistical data analysis. Quite unexpectedly, smaller plot sizes (15 square feet) were considerably more effective than larger plot sizes (30 or 45 square feet) at creating more uniform conditions. Incomplete block designs were highly effective, using discrete blocks in the field that contained only a subset of treatments. Sophisticated statistical analyses of the data were also effective at creating post-facto uniform conditions, particularly for the larger experiments. The methods described in this paper could be used by any researcher, who has several years of research history at a particular site, to improve the efficiency and effectiveness of field trials.
Technical Abstract: Field-based agronomic and genetic research is a decision-based process. Many decisions are required to design, conduct, analyze, and complete any field experiment. While these decisions are critical to the success of any research program, their importance is magnified for research on perennial crops due to multiple years of data collection. The objective of this paper is to summarize 28 years of field-based perennial forage grass research at a single location, describing changes to experimental design methodology illustrating both predicted and empirical results of those changes. The study is based on an analysis of total forage yield for 114 genetic experiments of 11 forage grass species. Over the course of time, plot sizes were reduced from 5.6 to 2.8 to 1.4 m2, resulting in a decrease in mean CV from 18.6 to 13.3 to 11.5%, respectively. These changes in precision, directly opposite that predicted from Smith’s Law of Heterogeneity, were attributed largely to a vastly improved relative efficiency of blocking and spatial adjustment as plot size was decreased: 212 vs. 130% relative efficiency of blocking and 240 vs. 109% relative efficiency of spatial adjustment for 1.4 vs. 5.6- -m2 plots. These changes suggested that spatial variation at this site consists of fine-scale variation that is uneven, unpredictable, and cannot be easily captured by incomplete blocking or spatial analyses of the larger experimental units. Finally, a power analysis was used to predict the number of replicates required to detect expected differences for a series of experiments, resulting in a high level of predictability and a highly successful application of power analysis to assist with the design of field experiments.