Technical note: assessing the consistency of measurement procedures in animal energetics and nutrition

Our objective was to assess the consistency of representative digestion and energetics determinations used in animal nutrition. We used distribution theory of quadratic forms that allow for the attainment of width of confidence intervals (WI) for intraclass correlations. Three models commonly used in animal nutrition were analyzed, and their respective programs were coded to obtain the required confidence limits. Data sets were obtained from previous research published by our laboratory. Urinary, CH4, and ME were analyzed assuming a two-factor nested balanced variance component model. Rate of ruminal NDF disappearance (kd) and DM digestibility by an 8-d conventional collection trial were fitted to a two-factor crossed variance component model without interaction with a single observation per cell. Empty BW (EBW), carcass energy, and EBW energy were fitted to a two-factor crossed variance component model with interaction. Widths of confidence intervals varied with the example data set and variable tested. The narrowest WI was that of DM digestibility, less than .07 at a 95% confidence level for all the intraclass correlations, which shows the high consistency of the DM digestibility measurement in the specific study. Medium to large WI were found for kd and EBW; WI estimates were less than .70 at a 95% confidence level. Large WI, from .8 to 1.0 at a 95% confidence level, were found for the remaining variables, indicating the greater variability of these measurements. This methodology allows the assessment of the consistency of a measurement process and provides a method to monitor it each time a determination is made.