SPATIAL VARIABILITY IN FIELD TRIALS

Abstract

Field trial experiments use blocks to improve precision. There are methods to adjust for different types of soil variation but they are not used due to ignorance or uncertainty about their capacity to improve the precision of blocks. In this work eight methods alternative to blocks were compared with regard to modeling the spatial variability . The results of 15 field trials with wheat (Triticum aestivum L.) and six with corn (Zea mays L.) from different locations in the world, were used for this purpose. The mean square and degrees of freedom of the error, the value of the F statistics for the test of equal treatment means, and the agreement of the residuals with the model assumptions of normality, independence and homocedasticity were used as criteria for this comparison. The polynomial and spatial analysis models were better than the others since they had more capacity to detect significant differences in the F test, and because they had residuals which were closer to the model assumptions, in most cases. The model proposed by the research workers, of incomplete blocks, was found to be superior to the other models only in one experiment.