Скачать с ютуб P- Hacking & Pop Science -- How to Prove Anything в хорошем качестве

P- Hacking & Pop Science -- How to Prove Anything

Statistical significance is usually defined as a “P” value less than 0.05, meaning that the difference observed between two groups would not be seen even 1 in 20 times by chance. That seems like a pretty high bar to prove the difference is real. But what if 20 comparisons are done and only the one that looks “significant” is presented? Or what if a trend is apparent in the data and samples are dropped or added to achieve this magic number of 0.05? And what if none of this is apparent in the publication and raw data are not available to allow for an unbiased analysis? Welcome to the world of "P-hacking” EAMES Consulting https://consulting.barryeames.com