Data Dredging

• rerunning tests increases Alpha error probability
  • Popper Logic
• also known as “data snooping” or “p-hacking”
• possible remedy: z-line
  • plot all results (even across papers) on a line, it should be normally distributed, otherwise it is biased
• examples, further details: https://en.wikipedia.org/wiki/Data_dredging