Plant breeders face the challenges to incorporate significant developments in Bayesian quantitative trait locus (QTL) analysis into breeding practice. The overall objective of this dissertation is to integrate QTL analysis into marker-assisted selection (MAS) in plant breeding using Bayesian statistics. Three different perspectives were studied: identification of optimal designs to generate multiple families for QTL mapping, cross prediction using QTL analysis results, and genomic selection for cultivated barley. First, the impact of two mating designs was studied on QTL mapping in multiple families generated by crosses between multiple inbred lines, using within-family linkage disequilibrium (LD) with a Bayesian variable selection approach. The loop design was found to have smaller mean square error in estimating QTL allelic variance and position. Second, the usefulness of crosses in developing inbred lines was investigated, using QTL mapping results from Bayesian shrinkage analysis. The usefulness of a particular cross depends on the expected performance of its best progeny, which was called the superior progeny value here. Theory was developed to predict the superior progeny value as a function of the mean of the breeding values of all progeny and of the standard deviation of the breeding values among progeny from a specific cross. Little difference among crosses for the standard deviation among their progeny was found under an additive genetic model for a trait, such that a benefit from estimating that standard deviation occurred only in relatively few cases. Finally barley marker data from 1803 SNP was used to evaluate genomic selection for breeding populations derived from 42 spring two-row barley lines. Three different genomic selection methods, random regression best linear unbiased prediction (RR-BLUP), Bayesian shrinkage estimation, and Bayes-B, were compared. The current barley SNP density was found to be high enough for genomic selection to better predict the breeding values of double haploid progeny than phenotypic selection. Overall, the Bayes-B approach that fitted a relatively high proportion of markers into the model had more stable performance across different scenarios. MAS for quantitative traits in plant breeding seem promising by integrating advanced Bayesian QTL analysis.