The objective of this work was to study different competition covariates in neighborhood analysis and to check the efficiency of this analysis in the increase of the experimental precision and the consequences in the ranking of sugarcane genotypes in terms of productive potential. Data used were from the state experimental network of the sugarcane breeding program installed in 2002, of the Instituto Agronômico, Campinas, SP, Brazil. The experimental design was in randomized complete blocks, with three replicates. The plots were formed of five 8 m rows with 1.5-m spacing between rows. The stalk productivity (TCH, Mg ha-1) was evaluated in the first and third harvests. The covariates "side", "tip" and "4 neighbors" were included in the Papadakis method and in two variations of the moving average method. The Papadakis and moving average 1 methods did not change the experimental precision in comparison with traditional analysis. The moving average 2 method reduced the environmental variance. However, it changed the estimates of the genetic variance and the genotype ranking, and could lead to erroneous conclusions in the selection. Although the "side" covariate was influential on competition, its effect or on five row-plots was small.
Saccharum; covariance; moving averages; Papadakis; experimental precision
