Hacker News Reader

This kind of thing is not science, but digging through data to look for “associations”. There are associations in any big dataset. There is no reason to infer from this that there is any cause-and-effect relationship whatsoever. But it can be an amusing exercise to think up confounders. Are patients with incipient dementia advised by their doctors to abstain from alcohol? Probably.

View on HN

While that's true often these types of weak observational studies are needed in order to fund more robust randomized and controlled studies which tend to be far more expensive and need some even low-quality evidence to justify.

In this case the evidence is particularly weak as even the paper notes almost every attempt at retrospective randomization or control makes the association basically disappear.

Reply on news.ycombinator.com