Use of linear discriminant analysis to characterise three dairy cattle breeds on the basis of several milk characteristics

Abstract

Listen

Translate article

To characterise individuals of differents breeds on the basis of milk composition and to identify the best set of variables a linear discriminant analysis (LDA), on 14 milk production traits, was performed on milk samples from 199 cows of different breeds (respectively, 127 subjects were Italian Friesians (IF), 62 were German Friesians (GF), and 10 were Jerseys (J) and all came from the same breeding farm in Tuscany. The variables were: test day milk yield (kg milk), % Fat, % Protein,% Lactose, % solid non fat (SNF), % total solid (TS), pH and titratable acidity (TA); five rheological variables: r, k20, a30, a45, and somatic cell counts /ml (SCC); and one hygiene-related variable: total bacterial count (TBC). The analysis performed on the 14 variables, with regard to the three breeds, allowed us to identify 10 of these as variables useful for discrimination (leaving out kg milk, pH, a45, and TBC). The most important variables were the percentage of Fat and TS for the first canonical variate and SNF, Lactose and Protein for the second. Fat and TS play an important role since they present significant values (even if opposite sign) in the two variates. The resulting classification of subjects was satisfactory: 79% of the Italian Friesians, 73% of German Friesians and 100% of the Jersey cows were classified correctly.