Best Linear PredictiOn: Further Topics

Abstract

Throughout the case studies of BLP in Chapters 5 and 6, a number of simplifying assumptions were made: first moments (E(g), E(y)) known, second moments (variances and covariances) known, observations in y restricted to family means from half- or full-sib progeny tests, all parents unrelated, only one trait measured (although possibly at different ages), and one trait predicted. Also, generally we assumed that the measured traits had homogeneous variance structures. Often in actual applications of BLP in genetic improvement programs, some or all of these assumptions may not be valid. The objective of this chapter is to show how BLP can be applied to broader, more realistic situations.