A model of discriminant analysis on the basis of descriptor variables for the ampelography of <i>Vitis</i> sp.

  • A. Camussi
  • A. Calo
  • A. Costacurta
  • C. Lorenzoni
  • E. Ottaviano
Keywords: ampelography, shoot, leaf, berry, biometry, analysis, descriptor variables, multiple discriminant analysis, genotype by environment interaction, normality assumptions


Use of descriptor variables in ampelography is recommended to simplify recording of data and to enable useful comparisons. Parametric assumptions are, however, poorly satisfied especially with regard to statistical interference. In the paper some statistical procedures to improve the discriminant ability of descriptor variables are considered. The use of variances and covariances of variety by year interactions is suggested for the error matrix within a multiple discriminant analysis procedure. The adequacy of this model is verified in a 3-year experiment with Italian wine varieties. The discriminant power, as evaluated on the basis of the estimated distances among varieties, is satisfactory.
Section 1: Genetic resources, evaluation and screening