RobertR
Senior HTF Member
- Joined
- Dec 19, 1998
- Messages
- 10,675
Further, if one of the amps dies because it can't handle the load, the test is scrapped.I thought the preconditions prevented this from happening.
Further, if one of the amps dies because it can't handle the load, the test is scrapped.I thought the preconditions prevented this from happening.
Should the formula for 240 iterations be "the same" ie
x = z_p^2*240/(z_p^2+1)
I would have expected the number of correct answers to decrease (per centage wise) as # of iterations increased.
Yes - it should be the same, but the original formula was incorrect (Dropped an N in the algebra )
The corrected formula is:
x = (n+z_p^2 + z_p*sqrt(n+z_p^2))/(2*(1+z_p^2/n))
This has the asymptotic behaviour that you are alluding to - keeping z_p low in proportion to n, x tends to n/2 as n gets large, thus a lower proportion of correct answers is needed.
My other post has been edited.
Aggregate results would be very interesting - although each test may be with different equipment, inference could still be made if the proportion of success is high enough.
I'll be doing this test myself at my house this weekend using my wife as the subject.If it's just you and your wife involved in the test, it will most likely be single blind, not double blind, which will leave enough potential for bias to invalidate the results as adequate evidence of differences.
What is up with all these statistics?The brief answer: Statistical Analysis
Regards,
My 20 years in pro and home audio (14 album credits) tell me that this is simply not true. Science cannot explain everything. Texture, tonality, and midrange purity as well as soundstage are not explained by any amplifier stats I know of. I wish it were that simple, it would simplify my purchasing greatly. I have no bias against science either as I operate a small AI software company as my day job. The best combination in AI is a blend of "art and science".Science can explain all of these things. It's just that our science isn't advanced enough to make the measurements that are required to explain the differences. We live in a world with rigid physical laws (above the quantum level) but we just can't model all of them due to the complexity.
If Richard Clark made the results of the 2000 participants available, then we could all see if there is any audible difference between the amps that were tested. Just because someone fails to get 24/24 doesn't mean there is no difference, all it means is they couldn't distinguish 100% of the time. I'd be highly surprised if people could distinguish many only slightly different things (Coke/Pepsi, Brands of Peanut Butter etc.) 24 times in a row!Blah, blah, blah...
So if I take two different color squares one red and one blue and alternate them in front of you in a similar fashion to a DBT asking you which is which you wouldn't get 24 out of 24 right? That would be unreasonable? You would only need to get it right 17 times to be statistically significant?
I think not.
If the differences in amps are anywhere near as large as the boutique amp makers would have us believe 24 out of 24 should be no problem. Especially for the people whose wives can tell the difference from the other room.
Seth