DC-2 vs. DC-1 v3.1

Posted by Jim Staszak [IP: 165.89.84.249] on August 25, 1999 at 09:09:01
Using Mozilla/4.5 [en]C-BMY (WinNT; U):

In Reply to: DC-2 vs. DC-1 v3.1 posted by Philip Brandes on August 25, 1999 at 01:10:00:

Philip,

Thanks for your thoughtful words.

I never meant to suggest that double-blind testing is the only touchstone of worth. It is cumbersome for the average person to employ. In my case I find methodological problems with most ABX comparisons. Since I am a statistician by profession, I'm sensitive to the violation of the assumption of statistical independence that is committed in many of these trials published in the audio press where the subject is not the unit of analysis, and thus the binomial probability distribution can no longer be relied upon. Secondly, when you're testing the null hypothesis of no difference and you fail to reject the null, you had better have had sufficient statistical power (statistical power is rarely stated in these comparisons). You can always fail to achieve statistical significance because of insufficient sample size. If the investigators are testing the null hypothesis of "no difference" and the intent is to demonstrate differences, that is reasonable. However, if the intent is to demonstrate "equivalence", then they are testing the wrong null hypothesis.

I am sincerely interested in any evidence that you have which suggests that ABX testing per se obscures real differences. I have to keep my mind open (it's difficult) to these things. But we can't argue that this proposition (ABX obscuring) is true or likely just because it is conceivable. What is the mechanism that governs this obscuring? It should then translate into a testable hypothesis.

I'm afraid that we are at risk of making the " perfect" the enemy of the "good by enumerating all of the difficulties of ABX testing " (e.g. 0.1 dB matching) and hence throwing the baby out with the bath water. Science doesn't deal with certainty, only with probability. I think good statistical and experimental design can address many of these issues.

Regards,

Jim

Follow Ups:

Return to the new SMR Forums Menu

Design & HTML © SMR Home Theatre, Images © SMR Home Theatre cannot be reproduced without permission.  The images on this page are digitally watermarked.  New forum messages should be posted into SMR Forums v2 - http://www.smr-forums.com/

	The Web   SMR Archive