Hacker News Reader

Comment by kelseyfrog | Hacker News Reader

Nothing? This is worse than nutrition studies in terms of being able to draw conclusions.

An RCT(randomized controlled trial) is the minimum standard of evidence necessary in this domain to perform a Bayesian update. Unfortunately, and even worse than nutrition, everyone is more tempted to weigh in with anecdotes, conclusions from bias, and arm-chair sociology.

At a minimum conclusions based on correlational studies like this should be disregarded. They're inconclusive at best and re-enforce already concluded preconceptions at worst.

View on HN

> An RCT(randomized controlled trial) is the minimum standard of evidence necessary in this domain

OK sure, I've randomly assigned you to have exactly 15 sexual partners before meeting your spouse at age 30, cohabiting for 3 years without having sex, then getting married with 100-150 guests at the wedding, having a child after 1 year of marriage and a second child 2 years after that. Also I'm going to need you to become Asian and devoutly Jewish, if you're not already.

Please report back in 1-2 decades to let us know how your marriage quality is.

That's rather generous of you and also a great example of why some knowledge is ethically and practically unknowable.

Rather than being ok with a rather uninformative conclusion though, we often see folks who cannot resist the part of their brain that desires clarity. For many different reasons, the discomfort of not being able to know, justifies reaching conclusions at the expense of them being accurate. So much so, even, that their accuracy is assumed rather than concluded.

Welcome to Sociology, see also Early Child Development.

After you have attended the year long cultural workshop on defining “marriage quality” with your peer group and wider community.

An extremely subjective topic with very controversial areas of discussion.

> Nothing? This is worse than nutrition studies in terms of being able to draw conclusions.

This research result doesn't have to prove causality conclusively to be useful.

It is an established fact that higher numbers of premarital partners among women correlate positively with likelihood of their eventually seeking a divorce from their husband. I think it's also been shown that this correlation exists for men as well, though it is not as strong.

This correlation is either causal or it is not. That is to say, (1) more premarital partners for women causes divorce odds to shoot up, or (2) women who have tendencies to have more premarital partners have some other quality that causes them to be more likely to get divorced.

Either way, good advice for a future husband is to avoid marrying a woman who has a lot of premarital partners (provided he wants to avoid being divorced).

The research result is useful for men to think about in spite of its inability to firmly establish causality.

Sorry, but it has exactly as much influence on behavior and decision making as nutritional studies do in advising people what to eat and how to exercise. ie: zero. Predictions based on correlations don't care whether anyone changes their behavior or not.

There isn't a special carve out for this topic in the realm of epistemological consistency.

I think I broke your brain.

There are no RCTs to prove that using a parachute leads to better outcomes than not using them when jumping out of a plane. Does that mean you wouldn't use a parachute if in that situation?

In case others aren't familiar with the study[1] you're referencing.

I'd be happy to jump from a plane in the same conditions referenced in the paper.

1. https://www.bmj.com/content/363/bmj.k5094

Reply on news.ycombinator.com