The aims of this study were to estimate the multiple linear regression equation and to verify the possible relationship between dependent and independent variables. Dependent variables were the mean percentage of the least significant difference (LSD%) and the mean square of the error (MSe) for grain yield. Data from 522 experiments conducted from 1986 to 1996 and 88 experiments conducted in 1997 were used in a total of 610 experiments of the National Competition of Maize Cultivars. In the 522 experiments, two regression equations validated by the analysis of simple regression between the real values and the foreseen for the equations were estimated, in the 88 experiments, it was observed that the regression equation was not a good estimation for the same original value for LSD%, but the equation can be used for the estimation of MSe. The application of Lilliefors test resulted in normal pattern distribution of MSe values. One classification table of MSe values was built based on observed values of variance analysis of the experiments and on the regression equation estimated value.
experimental precision; regression analysis; cultivar competition; Zea mays
