"Why Most Published Research Findings Are False" by John P. A. Ioannidis describes a statistical test for the likelihood of research being false, first without researcher bias and then a second test that includes bias. The result: it doesn't look good.