1. Randon sample
2. Statistically significant representative population
3. Causality shown
At least some of these studies have been very well done, and the results are overwhelmingly statistically significant.