Sufficient number of replications for path analysis in maize

The number of replications is assumed to interfere in the estimates of the path analysis coefficients. Thus, the objective of this work was to determine the sufficient number of replications for the path analysis of traits in maize cultivars. An experiment was conducted with 15 maize cultivars in a complete randomized block design with nine replications, and seven variables were measured. Then, 511 data files (matrices) formed by all combinations of the nine replications were organized, in groups of 1, 2, 3, 4, 5, 6, 7, 8 and 9 replications. In each matrix, containing the averages of 15 cultivars for the seven variables, Pearson's linear correlation coefficients were estimated, the multicollinearity diagnostics and path analysis were performed and dispersion diagrams were constructed. The sufficient number of replications for the path analysis was determined from the parameter estimates of the quadratic response plateau model. With the replications number increases, the accuracy of the path analysis coefficient estimates improves, but the gains in accuracy gradually decrease. Six replications are sufficient to perform the path analysis of agronomic traits of maize cultivars and can be used as a reference for designing future experiments.