Intro: p-values

• Originally: to indicate a problem with a model worth investigating. Not: p small therefore other model correct.
• p-values are not probabilities
• p-values are random variables between 0 and 1
• if the null hypothesis is true
• So the p-values should rarely be at one extreme
• If try often enough, it will happen
• The exact value of p-values below 1e-6 is unreliable
• don’t compare p=10-200 and p=10-100
• Effect size may be more interesting

http://astrost.at/istics/minimal-statistics-checklist.html