Data Augmentation and Complex Traits

The aim of livestock breeding plans is to improve an objective of selection by acting on several criteria of selection. The criteria of selection are composed by a subset of traits selected by its easiness of measure and its correlation with the objective of selection. In general, traits included on the criteria of selection are direct measures of the performance of candidates to selection. From an alternative point of view, the Bayesian paradigm allows to model the performance of livestock under a hierarchical Bayesian scheme. This approach can use the statistical technique of Data Augmentation, which allows to generate “pseudo-phenotypes” to mimic some complex traits. Later on, these “pseudo-phenotypes” may be used as new criteria of selection. The description of the procedure is illustrated with several examples of animal growth, ranking traits, subjective scores, food efficiency and disease resistance in several livestock populations. BAYESIAN ANALYSIS AND DATA AUGMENTATION The main objective of livestock breeding plans is to improve the performance of the livestock populations by acting on several criteria of selection. The criteria of selection are composed by a subset of traits selected by its easiness of measure and its correlation with the objective of selection. In general, traits included on the criteria of selection are direct measures of the performance of candidates to selection. Once traits included in the criteria of selection are identified, the Best Linear Unbiased Predictior (BLUP) usually obtains predictions of the breeding values after assuming the following linear model: y =Xb+Zu+ e where y is the vector of phenotypic records, X and Z are the incidence matrices that links with systematic (b) and additive genetic (u) effects and e is the vector of residuals. Linear models for genetic evaluation can be understood also under the Bayesian paradigm. Then, the Bayesian likelihood of data is: p y b,u,σ e 2 ( ) = 1 2πσ e 2 ( ) 0.5 e − yi−xib−ziu ( ) 2 2σ e 2