Multivariate nonnormality and multicollinearity in path analysis in corn

The objective of this work was to evaluate the effect of multivariate nonnormality and multicollinearity in the path analysis of corn. We used data from 13 corn cultivar competition trials. The response variable (grain yield) and seven explanatory variables (number of days to tasseling, plant height, ear height, relative ear position, number of plants, number of ears and prolificity) were measured in each cultivar. Then, data transformation and the univariate and multivariate normality diagnosis were proceeded. The correlation coefficients were calculated and the diagnosis of multicollinearity was performed, before and after data transformation. The path analysis was done according to three methods: traditional; under multicollinearity (ridge path analysis); and traditional with variable elimination. Data transformation reduces the degree of multicollinearity and the variability of the direct effects, in the traditional path analysis with high multicollinearity. Multicollinearity exerts more impact on the estimation of the direct effects in path analysis than multivariate nonnormality. The traditional path analysis with elimination of variables is more appropriate than the ridge path analysis.