Technical note: Analysis of ordinal outcomes from carcass data in beef cattle research1

Abstract

Listen

Translate article

Beef cattle research commonly uses Yield grade (YG) and Quality grade (QG) as outcomes in nutrition and health experiments. These outcomes, as commonly reported and analyzed, are ordinal variables with an assumed rank derived from an underlying latent variable that may or may not be available for analysis. The objective of this study was to employ mixed-effects ordinal regression and approaches previously reported in animal science and veterinary literature such as contingency table analysis, mixed-effects linear regression, and mixed-effects logistic regression for the analysis of YG and QG data and to compare results with respect to statistical significance and estimated statistical power. Five randomized complete block design experiments were used for initial evaluation. Simulated data sets were used for evaluation of relative differences in statistical power. Scenarios were observed where all of the methods differed in estimate of effect and statistical significance. Power to detect an association was similar between studies under the scenario evaluated. Ordinal regression approaches provide an estimate of effect that can be used in subsequent prediction of performance, which is an advantage over contingency table approaches that only report statistical significance. Further, ordinal models do not require modification of the outcome variable as in logistic regression or assumptions regarding YG or QG distribution in linear regression, which are often not met. Researchers faced with analysis of YG and QG data should consider the use of ordinal regression, particularly with recent advances in statistical software packages capable of implementing this method for data within hierarchical models.