Really hard to follow. He speaks not that great for the speed he chooses. And I agree with the criticism about cherrypicking. Slides are too detailed as well.
@arturleperoke3205
Жыл бұрын
@PyData I am really unsure what to do with such contributions to baysian ab-testing .. there were some points in the presentation which are plain wrong from my knowledge, like 9:35 "A test has been running for a long time, but *remains* insignificant" => that is plain p-hacking! You do not run the test UNTIL stuff is significant: wait long enugh and the chance for becoming significant at one point is high. You actually should determine the size of your sample before, collect data and evaluate as soon as you hit your determined sample-threshold. So, if I am interested but unfamiliar in the bayesian approach, how can I be sure whether that part was correct?
Пікірлер: 2