To JimP,
I have really enjoyed this thread. I’ve enjoyed it because it relates to some work I’ve been doing and enjoyed it also because it parallels problems I have had to deal with when working with statisticians for the past 25 years. Those problems were as tusenfem has pointed out, are the difference (or knowing the difference) between correlation and causation.
I worked for one of the US automakers as an engineer with a knack for solving problems. I was moved into a position where my job was solving problems others couldn’t solve. The problem I had was with each new problem came a pile of data and some statistician’s analysis of the data along with correlation/causation numbers. Never on any of these problems was the causal system suggested to me by the statistician ever correct. To be fair to the statistician, if they were correct I would have never been given the problem to work on. Statistics are an important tool, but by their self are pretty useless; but when combined with the knowledge of physics and other sciences and the understanding of systems can be very powerful.
I started several times to reply to this thread with encouragement and suggestions but was never able to put it into something short and simple. Nereid’s post was better than anything I could put together:
Quote:
Originally Posted by Nereid
Better yet: without something we might call a model - of the physical connection/relationship/mechanism/whatever - there is no way to test the (observed) correlation!
And that's how you do science (well, one way); make a hypothesis, develop it to the point where it is quantitative and testable, ... and go test it!
But wait! There's a whole lot of nose-to-the-grindstone stuff buried here, like logic, clarity, numbers, equations, .... and, if you want your hypothesis to (one day) be taken seriously, consistency with well-established theories (of physics), ...
If you don't do this hard yakka, you might as well say it's all down to invisible elves (or their metaphysical equivalents); of course, that's soooooo much easier!
|
The hard yakka sounds bad but it is really the enjoyable part of the whole journey. Imagine how you felt when you saw the R^2 of .97, I know it must have been exciting. You are seeing something that just maybe no one has seen before. Can you imagine what it would feel like if you knew why the R^2 was .97; I mean the physics behind it all. It would mean that maybe; just maybe, you are on a road that nobody has ever travelled before. Worst case is it’s just another ATM theory that will get shot down in an instant. But it’s the feeling that makes it all worthwhile.
Jim