Unless I'm mistaken, correlated means there is a relation. We don't know what the cause is, but we can see there is a relation. Meaning they at least appear to have something to do with each other. If we could show they have nothing to do with each other, they would no longer be correlated.

The cause of that correlation may not be what we think. Your link of "spurious" correlations has many examples. I would suggest there actually is a link between those things, but it's a 3rd factor of some sort. For example, the correlation between suicides and spending on scientific research may be correlated via a connection to prosperous societies. A society with more money will spend more on science. And there's lots of research to show that wealth leads to higher instances of mental health issues. So the correlation is real, but it doesn't mean one caused the other.

All of which is my long-winded way of saying I think there's tremendous value in these kinds of studies.

As I have said already in another reply to that post, what I mean is not all correlations are worth discussing. Obviously it has value from a scientific perspective but that doesn't mean it'll lead to any useful conversation. In fact it generally just seems to lead to insinuations and generalizations about large groups with little to no real evidence.

>I would suggest there actually is a link between those things, but it's a 3rd factor of some sort. For example, the correlation between suicides and spending on scientific research may be correlated via a connection to prosperous societies.

Edit: Not quite. What the graph is saying is there's a direct correlation between the how much the United States budget allocates to science, space and technology from year to year and suicide specifically by strangulation and hanging. There may be some vague link between suicide and science spending, but this is saying something much more specific than that.

> Unless I'm mistaken, correlated means there is a relation

Well the short rude answer is you are mistaken. The longer answer is that some correlated things have underlying causal relationships: one causes the other, vice versa, or a third unspecified correlate is causative.

But see that word some ? It's important. Because some are entirely unrelated, and the word correlated is not co-related in a causative sense. You might find piracy corollelates with consumption of potato chips. It doesn't have to mean anything.

The relation goes to trend: up or down. Positive or negative correlation. Why? The "why" kind of related, is different. Correlation is observed relationship, not functional, actual, causative relationship until a mechanistc reason is found, or it's absence is understood, and it's just.. pirates and chips.

The problem here is english. "Related" has different meanings.

> Correlation is observed relationship, not functional, actual, causative relationship until a mechanistc reason is found, or it's absence is understood, and it's just.. pirates and chips.

Be careful though: causal relationships exist where they do, and humans may not be able to find them and thus classify a causal relationship as merely a correlation.

I think correlation != causation misleads a lot of people on places like Reddit.

Correlation != Causation is the midwit take, so very appropriate for Reddit

Correlation implies a relationship that should be investigated for causal agents or hidden factors.

> Correlation doesn't equal causation > > For example, here are some weird correlations of no real significance

This is an nonsense take on science. Your two statements are totally unrelated to one another and mean completely different things mathematically.

Those graphs with "weird" correlations were generated backwards. They had data and looked for lines that look similar. Scientific studies are generated forwards (usually, let's not talk about retrospective studies which tend to produce junk "science"). We have a hypothesis we manipulate it and we see if there is an association.

If you write down the math for these two scenarios they are completely different statistically!

In one, you're always going to find another pretty line to match if you have data. It means nothing!

In the other, the likelihood of two things being correlated in your experiment when they are totally unrelated to one another is very low (assuming your experiment is designed well). The question is generally, what is the path from the variable I manipulated to the effect, not "random stuff happens, lets forget about science and live in this cave now."

So please stop saying "Correlation doesn't equal causation" it's nonsense that "smart" people repeat and then link to that website without understanding what's happening.

That website doesn't debunk science, where we run experiments without knowing the results. It debunks charlatans and liars that say "X looks like Y so they must be correlated". That's not what we do in science.

Correlation doesn't prove that two things are related; it does suggest it though. And a lack of correllation doesn't prove the two things are unrelated.

So a correlation between general BJW and dishonesty might be because both phenomena have the same underlying cause. That doesn't render the research meaningless. On the contrary, the correlation means that something is going on, and we need to look into what that is.

Causation or nothing is a false distinction. Pointing to the existence of some correlations that are spurious does not mean that all correlations are spurious. Every causation results in correlation.

Spurious correlations are easy to find, particularly with 10 data points and hundreds of variables to choose from (lol), but multiple hypothesis testing reveals those as spurious easily.

And everything that is caused by something is also correlated, it just requires more time and research to figure it out.

Technically true, but not every weak correlation is worth talking about. I would have thought HN had more rigor than that. If you guys want to talk correlations, there are much more interesting ones like 'Internet Explorer marketshare is positively associated with murder rate'.

Just because correlation doesn't equal correlation doesn't mean that correlation can't provide evidence that a model of behavior is correct.

When you can't measure something directly you make predictions of what it causes and then measure them. If that occurs it provides evidence that model is correct.

That doesn't mean that correlation means nothing though. It might mean something, or it might not. Correlation studies can give pointers as to where to concentrate more research efforts. That's valuable because money doesn't grow on trees and research costs money.

And yet, all causal relationships correlate!

Please stop using this idea to 'debunk' science. It does not mean what you think it means...