"Design falsification tests before claiming a finding is real When Patel found a spike in traffic fatalities on album release days, the team ran multiple falsification tests: they randomly selected 1,000 sets of placebo dates and found the real effect was larger than chance in 98–99% of iterations; they also applied each album's calendar date to years when no album was released and found no spike. As Patel explained, 'our real effect was larger than what you'd get by chance alone in 98% to 99% of iterations of this placebo test.' This process of actively trying to break your own conclusion — before publication — is the difference between a finding that holds and one that collapses under scrutiny. Anyone presenting data-driven arguments should run at least one deliberate attempt to disprove their own conclusion before declaring it real. Source: Vishal Patel, Freakonomics Radio, Smartphones, Online Music Streaming, and Traffic Fatalities"