Our models suggest that it is better to give large pots of money to a few groups, which can use the money to run studies with more data, than to give small pots of money to many people who can each gather only a few data points. The latter distribution is much more likely to generate spurious results for the propagandist.
