You want to impress your friends with detailed knowledge of the Duke breast cancer disaster?
Keith Baggerly and Kevin Coombes have collected everything you need to get started here:
- a video of one of Keith Baggerly’s talks (given at my institute in Cambridge),
- the 60 Minutes episode and clip,
- slides from a recent presentation with some of the newer details,
- the 2009 Annals of Applied Statistics paper,
- an editorial they wrote for Clinical Chemistry about what information should be required to support clinical “omics” publications, and
- links to the Institute of Medicine session recordings and slides.
Systemic problems are more important than fraud
In an email Keith sent out to his ‘spam list’ today, he explains why fraud is not the whole, or even the most important, part of the story:
While some of the errors might be due to fraud, we’ll bet some of the errors were completely inadvertent (off by one indexing of gene labels comes to mind). (…)
Two bigger problems we encountered were
- that raw data and code were often not supplied, and
- that even when we identified problems with the data and sent reports and code illustrating these problems to various parties (the authors, journals, the initial Duke review), these parties dismissed the warning without checking the data themselves.
These problems were mentioned, but only in passing, in the 60 Minutes segment.