[omega4412]

Actually, peer review, by design, CANT'T detect fraud. Peer reviewers do not attempt to replicate what the author of a study did.

You're right that reviewers don't try to replicate the results, but I wouldnt say CAN'T detect fraud. Reviewers and editors have found problems in figures that turned out to be fraudulent. See Nature 439, 520-521 (2 February 2006) "Forensic software traces tweaks to images"

[curiosity]

Yeah, I guess you're right. They look carefully at descriptive statistics and the like, and sloppy fraud can often be exposed simply by finding inconsistancy in these things.

But if the fraud artist is smart, he can ensure that such things are not going to show up.