Effect of a pen in group-feeding trials: modelling data from suckler cows

Abstract

SUMMARYThis paper provides information on the extent and containment of intraclass correlation (ICC) concerning group-feeding trials on suckler cows. The research comprised the re-analysis of six previously reported experiments and a parallel simulation study. ICC coefficients were estimated for eight variables. They were seen to be higher in experiments on bulls compared with those on cows and more notable in the measurements of live weight gain compared with final weight. Moreover, the ICC coefficients were generally high for all variables measuring behavioural patterns. The simulation study showed that using a single animal as an experimental unit could be valid in certain situations, but it could not be extended to cover all the cases. The simulation study also showed that the common mixed model approach had significant problems when the ICC was slight. In general, the research strengthened the arguments that much more effort should be placed on the planning and statistical analysis of group-feeding experiments, especially in behavioural studies.