It can work for e.g. theoretical physics. But for anything involving serious extermination you've got to take the word of the researchers that they honestly did and reported it. At the most extreme example, experiments done with the CERN Large Hadron Collider are not easily replicated elsewhere at this moment ^_^.

It could also work for the explanation parts of experiment based papers, but, eh, you're allowed to be wrong there. Just publish enough of the data allowing others to double check your interpretation.

It is left as an exercise to the reader what other fields are known for withholding data, and the likely reasons why.