It leads to a definition of the adjusted P value, which is the answer to this question:The problem is a form of multiple comparisons.Many studies are done, but only some are published, and these are selected because they show "desired" results.Statistical analyses can be interpreted at face value only when all steps are planned, all planned analyses are published, and all the results are considered when reaching conclusions.These simple rules are violated in many ways in common statistical practice.If you try hard enough, eventually ‘statistically significant’ findings will emerge from any reasonably complicated data set.data tortureP-hacking(9)When reviewing results, you often can't even correct for the number of ways the data were analyzed since the number of possible comparisons was not defined in advance, and is almost unlimited.When results were analyzed many ways without a plan, the results simply cannot be interpreted.At best, you can treat the findings as an hypothesis to be tested in future studies with new data.