Home Propeller Head Plaza

Technical and scientific discussion of amps, cables and other topics.

RE: Searching for truth?

Rick,

I don't necessarily agree with the author that if you don't know how to improve the detectability (SNR) then the hypothesis probably not scientific.

As a counter-example, where would particle physics be if we had rejected the standard model back in the 1960s on the basis of warning sign #3 - that many of its predictions were essentially unverifiable with known technology and involved nearly undetectable weakly interacting particles? Or if we had rejected it on the basis of warning sign #7? Instead, we invested heroic amounts of money and effort trying to confirm the theory's predictions, and it turned out to be right.

I agree that a good first step to tackling a problem is "observing the problem while futzing with as many variables as possible to ferret out those that affect it." That's how I usually start troubleshooting. However, I don't think it's nearly as effective at optimizing. When troubleshooting, you are starting from having observed a problem and it's usually a pretty easy observation to repeat and confirm. But when you experiment to find improvements, I find the observations to be a lot noisier and the variables are not always independent.

And then you have to deal with patternicity - human nature seeking explanations for patterns that are essentially random, or at least caused by some uncontrolled variable that we weren't interested in. This seems to be a common problem in machine learning and other fields involving statistical classification, where the classifier adapts to features evident in random data which are not caused by the underlying function/mechanism of interest. There is a similar tendency in human perception, which evolved to evade predators and hunt prey. Our minds are well adapted to identify patterns in sensory data and then learn to recognize them. It seems like we're hard wired to to over-classify; we tend to see and interpret patterns even in random data, and don't easily recognize randomness unless we're looking at a sufficiently large data set where the distribution is obvious. I think this is part of the reason why there's a lot of tail chasing in audiophiledom. When you approach the optimum then it becomes harder to control variables and harder to separate an effect from noise, and you can get off-track chasing meaningless patterns in noise.

The only way I know of to avoid tail chasing is to make a lot of observations over a long period before making any conclusions.

Dave


This post is made possible by the generous support of people like you and our sponsors:
  Crux Audio  


Follow Ups Full Thread
Follow Ups

FAQ

Post a Message!

Forgot Password?
Moniker (Username):
Password (Optional):
  Remember my Moniker & Password  (What's this?)    Eat Me
E-Mail (Optional):
Subject:
Message:   (Posts are subject to Content Rules)
Optional Link URL:
Optional Link Title:
Optional Image URL:
Upload Image:
E-mail Replies:  Automagically notify you when someone responds.