| The focus of this dissertation was the development and use of semi-parametric methods for statistical learning in two areas of application: quantitative genetics and production economics. A common motivation in both cases was to develop and evaluate methods that, making use of available knowledge, have the flexibility of capturing complex patterns between the variables being studied.;Applications in this dissertation included: (a) a modified version of the Bayesian LASSO (least absolute value shrinkage and selection operator) for prediction of genomic values using dense markers and pedigrees, (b) models for non-parametric prediction of genetic values using reproducing kernel Hilbert spaces methods, and (c) a semi-parametric stochastic frontier model that allows inferring features of a technology and the technical efficiency of individuals firms using semi-parametric methods. |
