Applications of the principle component analysis (PCA) at grape varieties from the sortogroup Coarnă neagră for establishing phenotypical variability

Abstract

Multivariate analysis allows the analysis of variables of different individuals, data measured together. Therefore, o complete description of the interdependence relations that exist between groups of individuals and the measured variable is obtained. The multivariate analysis method deals with reducing the number of innitial variables by substitution with others resulting from their combination. This can be represented in a graph by points in a two dimensional or three dimensional space, while not loosing an excessive amount of information. The PCA technique can be seen from more points of view. For classical statistics, PCA is the determination of main axes of an elipsoid, indicator of a normal multivariate distribution, these axes being estimated as random samples. PCA is a graphic representation of these, having as optimal character according to some algebraic and geometrical criteria that does not presume the emission of a initial hypothesis of statistical nature on the data that is to be analyzed. PCA allows the extraction of the maximum information, in a simple and coherent form, as a data ensemble, by underlining the interrelations between variables and individuals, either by similarity or opposition.