Experimental precision of spatial analysis methods to evaluate the productivity of common bean families.

Abstract

The analysis strategies that model spatial dependence, in the form of covariance analysis, have been suggested in plant breeding, in order to exert greater local control and to increase experimental precision. This study aimed to evaluate the degree of spatial dependence of errors in productivity experiments to evaluate common bean families, to compare the experimental precision of the methods of spatial analysis, moving average and Papadakis, identify the most suitable neighboring plots for the calculation of the environmental index, besides proposing and evaluating changes in the application of these methods of spatial analysis. Data of grain yield of common bean families of the winter season 2006/2007 and drought season 2007/2008 were used. In these experiments, the number of families evaluated ranged from 25 to 400, characterizing the 5 × 5 to 20 × 20 square lattice. The moving averages and Papadakis methods, combined with the randomized block design, ensure spatial independence of the errors and presented experimental precision similar to that of the lattice analysis. For the calculation of the environmental index, the highest experimental precision was obtained from the use of two neighboring plots, one upper and one lower than the reference plot. The proposed modification in the application of the methods of spatial analysis provided greater experimental precision in relation to the lattice analysis, especially in experiments with moderate spatial dependence and using the reapplication of the Papadakis method