Published online by Cambridge University Press: 03 April 2017
The layout of experiments for the study of the growth of four crops in the central rainlands of the Sudan in 1965 is described. The treatments, different levels of flooding prior to sowing, were replicated within blocks. Standard procedures for estimating within plot, between replicate plotswithin block and treatment x block components of error are set out and an example is given of the analysis employing data on kenaf. Using typical values for components of error found in these experiments, increases in precision obtainable by varying the number of plants sampled per plot, and the number of replicate plots per treatment, are considered in relation to their costs. It is concluded that excessive replication would be required to establish differences of interest and that attention should be given to sources of error variation when such large errors are found.