One group of researchers thinks headline-grabbing scientific reports are the most likely to turn out to be wrong