Abstract
Research in the area of pastures, forages, and grazing livestock has a storied history within the realm of statistical analysis. Unlike traditional experiments in ruminant nutrition, in which an animal is fed individually and data are collected to assess the applied treatment, research on the grazing animal presents its own unique set of challenges. The collection of data on multiple scales (e.g., animal, pasture, landscape, time) brings into question the appropriate assignment of the experimental unit, and variance and covariance estimates must account for both the spatial and temporal effects of the environment. Oftentimes, the designs, assumptions, and rules-of-thumb taught to us in graduate school do not meet muster to adequately address the intricacies of this situation. This presentation will seek to address these complications and present statistically-sound solutions to obtain the maximum information from experimental data. First, a historical examination will be offered of how grazing experiments were originally handled. Next, conjecture will be offered as to why these methods may not remain valid and how advances in computing power and statistical theory allow us to obtain more information from the experiment. Finally, solutions to common scenarios will be offered whereby a more adequate or complete description of the experiment may be obtained.
