Evaluation of four alternatives of analysis of experiments in square lattice, with emphasis on estimate of variance component

The efficiency of fur alternatives of analysis of experiments in square lattice, related to the estimation of variance components, was studied through computational simulation of data: i) intrablock analysis of the lattice with adjusted treatments (first analysis); ii) lattices analysis as a randomized complete blocks design (second analysis); iii); intrablock analysis of the lattice with non-adjusted treatments (third analysis); iv) lattice analysis as a randomized complete blocks design, using the adjusted means of treatments, obtained through the analysis of lattice with recuperation of interblocks information, having as the residual mean square, the average effective variance of this same lattice analysis (fourth analysis). The results showed that the intrablock model, must be used with adjusted treatments in lattice experiments to estimate variance components every time that the relative efficiency of the lattice design, relatively to the randomized complete blocks design be upper to 100%, and in the opposite case the randomized complete blocks design model must be used. The fourth alternative of analysis must not be recommended in both situations.

lattice designs; variance componets; effciency; simulation