Published online by Cambridge University Press: 03 October 2008
In Ghanaian cocoa progeny trials plot values were adjusted by covariance on neighbouring plots. Standard errors of differences between progenies were usually reduced by 10% and often by over 20% when covariates were formed from the residuals of neighbouring plots and analyses were iterated using successive improved estimates of treatment effects to derive fresh residuals. The method gave larger reductions in error and was effective over a wider range of variates than adjusting on early growth data from interplanted temporary trees of a standard variety.