Yep this ironical article illustrates exactly the central defect in Cytodyn's paper that makes every data manipulated:
the complete lack of control for multiplicity for all the post-hoc tests. They apparently went on a fishing expedition, performing multiple tests on anything that could matter until they found the best possible magical combination (split by age at 65 years old precisely, no more no less, then split following the other treatments, and only some of them precisely handpicked). And of course they never discuss correction of the p-values obtained from that multiplicity of tests. Which is, statistically, completely illlegal and makes p-values completely fake.
We are not talking here about minor corrections: to make it correct again, we may need to multiply Cytodyn p-values by x30, x50 or more (we don't know how many tests they performed in their fishing expedition) to correct for this data mining. The resulting p-values are so irrelevant that it proves that Cytodyn has been just overfitting noise. It's a giant lotto game, they have thrown dozens or maybe hundreds of balls (the tests performed), who knows, and then, have handpicked only the balls that served their storyline. And have conveniently dissimulated the key statistics about the other balls from shareholders.
Market Data 
Markets 
AI
