Design and analysis of grazing and multi-tiered experiments
(C
Brien with C Demetrio and R Bailey)
It is often difficult to determine the appropriate linear model, and hence analysis of variance, to be used in analysing a grazing trial. This difficulty arises from the need to randomise both pasture treatments and animals to the plots. The derivation of the appropriate linear models, and associated analyses of variance to be used in analysing grazing trials shows how the randomisation that was employed in the experiment is incorporated into the analysis.
Multi-tiered experiments are experiments that involve certain forms of multiple randomisation. They include two-phase experiments and some laboratory, some superimposed and some grazing experiments. Different types of multiple randomisation that can occur in multi-tiered experiments have been identified. The implications of this for the structure on the factors, and hence for the analysis of variance decomposition, are being investigated.
In particular, an investigation is made of the conditions that must be placed on the designs employed to ensure that the analysis remains balanced.
More information is available from the multitiered experiments web site.