Data from 14 egg production experiments, which had replicated pens and individual trapnest records, were analysed to obtain estimates of experimental error from which to determine pen effects. The experiments were all nutrition experiments and were conducted at four experimental stations.Analyses were carried out on both a hen-housed and survivor basis. For all experiments the sampling error (individuals) underestimated the experimental error (individuals + pens) by about 20 per cent on a hen-housed basis. On a survivor basis there was no evidence of pen effects, that is, the sampling error did not underestimate the experimental error.The coefficient of variation for all experiments was 40 per cent on a hen-housed basis and 25 per cent on a survivor basis.The relative efficiency of two experimental designs, randomized block and completely randomized, was calculated for 11 of the experiments (on a hen-housed basis only). With the exception of 2 experiments, there was an increased efficiency of up to 500 per cent in utilizing a randomized block design. The data further indicate that the increase in efficiency obtained with a randomized block design is much more marked when the blocks represent different houses than when the blocks consist of different locations within a house.
A large-sample Kolmogorov-Smirnov test for normality of experimental error in a randomized block design