Use of multivariate discriminant methodologies in the analysis of phenotypic and genomic data of cattle

The present thesis is organized in 4 main chapters. The Chapter 1 is the general introduction and it regards the use of the multivariate statistical techniques in animal science, with a particular emphasis on the discriminant analysis.In Chapter 2, a new statistical method called Discriminant Association Method (DAM) was proposed. The DAM approach, developed by using multivariate statistical techniques, overcomes most of problems that affect the single SNP regression technique used in the ordinary GWAS.In Chapter 3, a new index to evaluate feed efficiency was defined: the residual concentrate intake (RCI). RCI identifies efficient and inefficient bovines in converting the concentrate. RCI can be quite simply evaluated and, in consequence, it could be easily included in genomic breeding programs. In the present research, the DAM method was applied to develop a GWAS for selecting markers associated to RCI.The research reported in Chapter 4 was aimed to develop an algorithm able to early identify highly persistent lactations. Four different models were fitted to individual lactations by using the first 90, 120 and 150 days in milking. Two multivariate statistical techniques were exploited: the canonical discriminant analysis (CDA) and the discriminant analysis (DA). The proposed algorithm combines the talent of curve models in depict features of the lactation and the ability of multivariate statistical techniques in distinguishing differences between groups.