[js1138]

Wasn't there a scandal a few years back when Newton was accused of "cleaning" his data to make it fit his laws. Seems he didn't have the statistical tools to deal with normal errors of measurement and was afraid to publish data that didn't fit.

[Doctor Stochastic]

Mendel was also accused of adjusting his data. The data fit too well. (My own guess is that Mendel didn't report data that failed to fit. The well known Desk-Drawer problem.)

Peer-Review is designed to catch errors. It's not as good at catching fakes, but generally it does so eventually. When I review, I generally duplicate all the computations in a paper. (I can't always afford to do the experiments.)