alt Hacker News>Couldn’t this effect be classic cause vs correlation?
Sometimes changing the correlated item, also affects the cause, through a link of causual changes.
E.g.: "Night visits to the fridge linked to high cholesterol".
Now, that's just correlation: it's not the visiting of the fridge, it's the snacking.
But if you read that and stop visiting the fridge, you likely reduce your snacking too as a side effect, and thus lower your cholesterol, without consciously trying to address the primary cause.
I feel like your example is flawed, I just can't put my finger on it.
Maybe it's because I don't see how sleep regularity is a factor you can change as willingly as visits to the fridge, or maybe its because I don't see why people wouldn't just eat more before heading to bed.
It could also just be that I find a treatment of symptoms to be less desirable than causes.