Statistical analyses for comparison of esophageal and hand-clipped samples from grazing trials

Abstract

Esophageal fistulated animals are used to collect samples of the forage being consumed by animals in grazing experiments. Four principal hypotheses (H) can be tested in esophageal trials: (1) esophageal samples are similar over treatments, (2) available forage is similar over treatments, (3) esophageal and available forage samples are similar, and (4) differential selection of diet did not occur among treatments. Because of the constraints of limited animal numbers, multiple latin-square or crossover designs in which the same animals are used to sample pastures during different periods of time are used to test H 1. Available forage is determined by collecting samples from the pastures over the duration of the study. The experimental design for these samples is a split-plot in time which is used to test H 2. Analyses of 1 set of samples (esophageal or available) using the experimental design for the other set is inappropriate. Since esophageal and available samples are paired within experimental units, paired t-tests can be used to test H 3 by treatment or averaged over treatments. H 4 can be tested by conducting an analysis of variance of esophageal minus available (or vice-versa) differences averaged over periods. Significant treatment effects indicate differential selection of diet among treatments occurred.