Quote:
There are many
published studies utilizing within-subject designs where a patient can receive several treatments (drug and placebo). Within-subject designs also have dedicated statistical analyses that (by eliminating between-subject variance) are quite powerful and
widely utilized.
Right on! Also called "SS1" or "single-subject designs". Sometimes the order in which the treatments are given matters, and then the analysis is "repeated measures design" or "individual as block" designs. Any modern text on experimental design covers these.
But there's more. A drug treatment may very well have a real and (somewhat) lasting effect on the subject. That's a problem in repeated measures design. Not the case with audio testing, although there are audio memory issues. There are also ethical issues in giving sick people "placebo" instead of "drug", so some designs try to minimize this. Not the case with audio testing.
The measurement in drug trials is not as subjective on the part of the subject generally (sometimes it is, as in "are you feeling less suicidal now?"), while in audio it is 100% subjective and subject to wicked bias. That's why false comparisons (swindles) of A/A and B/B are so powerful and important in audio ... they smoke out some types of response bias:
-- I love to show off, I want to help the tester, I tend to find differences even when they don't exist
-- I always love the second thing I hear ... it's better, right?
Having a known clunker C is also very powerful, smokes out:
-- I hate this nonsense, and the test leader is an ugly jerk, I ain't reporting no difference no how no matter what I hear
These are powerful biases, typically unrealized by the subject himself!
Wait, there's more!
The point of a drug trial generally is to establish a base level of efficacy in the population so the drug can be released. In audio, unless you are a manufacturer or a magazine doing research, the population effect is less interesting. Pooling the results of SS1 is not important
to me. I only want to know what I can hear and what I can't.
Borrow gear, use audition and return, buy-and-resell, go to meets ... do your own blind testing, typically single not double blind with an unbiased friend doing the gear switch, throw in false A/A and B/B comparisons, rotate in a known piece of junk C if you can ... use everything else (physical measurements, reviews, group blind tests) only as a rough guide. Be aware that non-blind testing will ALWAYS have the placebo effect on you no matter how hard you try.
Enjoy the music and don't upgrade so much.