Gelman: "a lot of studies just shouldn't be done". too much noise, too poor an experimental design.

#preregistration

Researchers need to start simulating what data they are expecting to collect before they collect it.