[timr]

Ya know, I get really frustrated whenever I see a bunch of armchair scientists work themselves into a lather criticizing the "sloppiness" of published research, based on nothing but a short blurb on a website.

Contrary to the belief of most indignant internet geeks, there's a lot more to scientific analysis than parroting "correlation vs. cause!" whenever data are presented. The fact of the matter is that, while correlation doesn't imply cause (that's the actual expression; it is rampantly mis-stated), a strong correlation is usually evidence that an underlying cause exists, whether direct or indirect in nature. Furthermore, there's not an accepted result in the history of science that doesn't depend on the analysis of correlations in observed data. Even falsifiability (that sexy buzzword of Popper-idolaters) depends on the collection of observations, and correlation of those observations with previous data. Thus, in a very real way, science is the study of correlations, and you can take the amateur correlation/cause pot-shot at any scientific result.

So, when a scientist points out that correlation does not imply cause, what she means is not that the observed relationship doesn't exist, but that the correlation could be caused by latent, indirect factors. This is a very different thing than suggesting that the result is invalid -- which is what most armchair, Popper-loving internet-scientist types are trying to do (including the parent posts).

[DanielBMarkham]

Sorry for the delay in getting back to you. My lather was starting to dry and I needed a good rinse.

One thing and I'll let you go. "a strong correlation is usually evidence that an underlying cause exists, whether direct or indirect in nature"

Oh really now? Well that explains my theory as to how carrots are deadly poisonous. You see, I've found a 100% correlation between the consumption of carrots and death, usually within 80 years. I'm also working on a theory that sunlight is a cause of all cancer, seeing as how there is another 100% relationship between all people with cancer and exposure to sunlight.

Science is the creation and study of models, abduction, based on the use of deduction and induction. Correlation is simply a tool used to create the model. When we forget we're comparing models, then it's easy to simply fall back on simple correlation-based arguments. That's like forgetting geometry and simply observing the circumference of a circle is usually about 3 times the diameter. It's true, but it's not creating an abstract calculus that can then be extended.

I find your continued use of "armchair scientist" to be an ad hominem, as if by using that name my argument would carry less weight. Last I checked we weren't providing credentials to express opinions.

I'm done playing. timr -- you can follow me around to another comment on the board and continue your picking. I have to go get my armchair adjusted for physics.

[timr]

Your examples bely either a fundamental misunderstanding of the mathematics of correlation, or a willingness to ignore them to fake an argument.

First, you can't "correlate" with incidence of death -- it's not even a binary variable, since everyone dies. Of course, you can correlate with onset of death, but that would make your example silly; if there were an observed acceleration in onset of death due to carrot consumption, it would only serve to illustrate the value of correlative studies in science.

Likewise with sunlight exposure -- you've just flipped the constant to the predictor (everyone has experienced sunlight exposure), and attempted to correlate with a variable output (cancer occurrence). Again, mathematical nonsense.

That said, I don't dispute anything you've said about models, abductive, deductive or inductive reasoning. But you can reason all day long, and without empirical data, you've got nothing but a castle on a cloud. Ultimately, you're always going to come back to correlating at least one observed variable with another -- the predictor, and the response. This is basic science.

Why do I refer to "armchair scientists"? Because I couldn't think of a better term to describe those people who want science to be just like math, but conveniently forget that mathematics rests entirely upon a set of fundamental axioms that are simply asserted to be true. The people who learned about one logical fallacy some time in high school, and latched on for dear life, conveniently forgetting that every scientist in the world knows about the same fallacy.

Moreover, these folks aren't familiar with the details, so they don't see that the places where mathematics most closely touch science -- quantum theory, string theory, etc. -- tend to be extraordinarily messy. The math frequently depends on assumptions and approximations that are accepted based on their similarity to observed data. For example, right now, thousands of scientists are banging subatomic particles together in a subterranean tube in France, in order to observe something that the mathematicians think might be there. If it isn't, the mathematicians get to start over.

Again: it all comes back to observing things. And as long as we're observing things that are difficult to observe, I have a feeling that cranky internet scientists will be there to argue from the comfort of an easy chair that empiricism has it all wrong.

[asdflkj]

How can a non-scientist better understand science? Do you know of any good books?