The nature of confounding in genome-wide association studies.

The authors argue that population structure per se is not a problem in genome-wide association studies - the true sources are the environment and the genetic background, and the latter is greatly underappreciated. They conclude that mixed models effectively address this issue.