2 ms·

If you have fifty different groups studying roughly the same thing and publishing any interesting-looking results with p<.05, you're probably going to have some

If you have fifty different groups studying roughly the same thing and publishing any interesting-looking results with p<.05, you're probably going to have some spurious results on the topic. If you have 500 groups, you're going to have a lot of spurious and probably some outright contradictory results published.

jaggederest 3y ago
In many cases, with large enough numbers, the spurious results will outnumber the real ones, especially so since true-null-hypothesis results are often unpublished.