There are currently huge amounts of data being generated on human genetic variation. For example, the recent International Hapmap Project generated data at millions of genetic markers for hundreds of individuals, and ongoing genetic association studies are typing thousands of individuals at 500,000 or more markers. These kinds of data possess a complex correlation structure, due partly to the (unknown) relationships among individuals sampled from a population. Developing accurate and computationally-tractable statistical models for these kinds of data poses an interesting challenge. We will describe some recently-developed models, and some of their applications, including estimation of missing data, estimation of fine-scale recombination rates, and gene mapping.