[boombip]

> This might also be another avenue for causation.

If people with mental disorders were more sleep-deprived and then consumed more sugar, wouldn't that also show up in the data?

> Although the study looked for this link, it was not seen in the data: men and women with mental disorders were not more likely to consume more sugar.

A causal A > B > C link would look similar (data-wise) to a A > C link.

[tylerhou]

Not if A => B for only a small population of A. But in that case, you still might see a C => A link.