Is it normal in a study like this to report the results separately for males and females? At low intervention levels what’s significant for males is not significant for females and vice versa, so there’s some potential for statistical shenanigans.
Agree with this too. On the one hand, Simpson’s Paradox, on the other hand, Simpson’s Paradox. But if you don’t expect male/female confounder, it just gives you three goes at the magic p<0.05.
Is it normal in a study like this to report the results separately for males and females? At low intervention levels what’s significant for males is not significant for females and vice versa, so there’s some potential for statistical shenanigans.
Agree with this too. On the one hand, Simpson’s Paradox, on the other hand, Simpson’s Paradox. But if you don’t expect male/female confounder, it just gives you three goes at the magic p<0.05.
Frequentism. It’s just broken.