130.164.67.107
In Reply to: RE: You underestimate the thickness of my skull posted by EBradMeyer on September 25, 2007 at 11:33:53
The second test I attempted was at home instead of at work, and it failed miserably (7/16 or something like that). That bugged the hell out of me, so several of the tests following that were simply attempts to isolate the system change that triggered the change in result. That chase alone resulted in an extra five tests.
The chase was not for naught: I couldn't get a good statistical result at all with one pair of headphones, and getting good results for another sound card was a hell of a lot harder. That could be the start of an extremely interesting test: One might be able to correlate the difference in results to some objective THD criteria.
After that, people commented that I shouldn't have run the tests without fixing the number of trials beforehand, because I could have cherry-picked where the test stopped. That alone necessitated redoing the tests at N=32.
It's worth noting that I've done a lot fewer ABX tests since I did that. :)
Follow Ups: