This work presents an alternative analysis of field experiments using geostatistics (correlated errors) coupled with conventional analysis (independent errors). The evaluation of this approach was based on the standard errors of the mean estimates for variable of interest (crop yield) from a field trial of soybean lines designed augmented blocks. The proposed method results in an average decreasing of 24.04% on the estimates of standard errors, which indicates that the method is efficient.
geostatistics; mixed model; semivariogram; spatial dependence