Most scientists agree: many published results are hard to reproduce. Can open source practices help fix the problem?