What statistical measures can you use to figure out if a correlation [like LDL] is active on one end of a distribution but not the other (eg u or _/-shaped curve)?

[AlexKChen]

#statisticalmethods thread, since there needs to be one on better statistical analysis here [with so many studies being misinterpreted]

_/ shaped curves are particularly indicative, the problem is that when you use the SAME correlation across the same population, you’re magnifying the effects for some subgroups and minimizing the effects of others

ultimately this also needs to lead to better discussion of control/cohort matching, mendelian randomization, and