What are the evolutionary and molecular causes of human phenotypic variation? Genomic data contains information to address this question, but extracting this information remains challenging. In the first part of the talk, I will discuss a statistical model for learning about the patterns of population splits and mixtures in the history of a species, with applications to humans and dogs. In the second part of the talk, I will discuss a statistical model for combining a genome-wide association study with hundreds of functional genomic datasets to identify the classes of genomic element for relevant for interpreting loci associated with a trait. I will show applications to 18 traits, including red blood cell traits, lipid levels, height, and Crohn's disease. In the last part of the talk, I will discuss extensions of the model to learning about pleiotropic effects of alleles on multiple traits, and highlight examples of unexpected connections between traits