[Oliver Costich]

On Fri, 18 Jan 2008 20:33:43 -0800, "JBorg, Jr."
wrote:

Oliver Costich wrote:
Mr.clydeslick wrote:
Oliver Costich wrote:




snip

Back to reality: 61% correct in one experiment fails to reject that
they can't tell the difference. If the claim is that listeners can
tell the better cable more the half the time, then to support that
you have to be able to reject that the in the population of all
audio interested listeners, the correct guesses occur half the time
or less. 61% of 39 doesn't do it. (Null hypothesis is p=.5,
alternative hypothesis is p.5. The null hypthesis cannot be
rejected with the sample data given.)

In other words, that 61% of a sample of 39 got the correct result
isn't sufficient evidence that in the general population of
listeners more than half can pick the better cable.

So, I'd say "that's hardly that".

you seem to be mixing difference with preference, you reference
both, for the same test.

For the purpose of statistical analysis it makes no difference.


But for the purpose of sensible analysis, shouldn't it makes a
difference.

I don't think so. I can't see any way the statistical analysis would
be different.

As you have said that logic is on the side of not making decisions
about human behavior. Isn't this reqiured to ensure sufficient testing
using well designed experiment and statistical analysis.

I didn't say that.


And just what is the general population of listeners.

You tell me. I presume that those who attend CES and would be
a good one to use.

That could very well include someone like Howard Ferstler, a raving
lunatic with a well-known hearing loss out to destroy high-end audio
and derogate all audiophiles young and young at heart. Provided,
of course, he can *afford* the fares.

Obviously you want to weed out people who are absolutely sure you
can't tell. But leaving out people who are skeptics biases the result
as well. I doubt that the 39 people who did the test comprised a
simple random sample, another design flaw. On the other hand I'd like
to see a well designed test using a simple random sample from the
population of true believers just to see if they can really. Even if
some people can tell, I suspect that it's a very small number. I do
know a couple of people who can really lock onto particular
characteristics and use then to identify what's playing.


What would you use and how would you construct a simple
random sample from it?

Are you testing the 99% who don't give a rat's
ass anyway? If so, so what. Or are you testing people who actually
care.


We need a bias controlled experiment.

Yes but the neither the golden ear cult or the nonbeleivers would
accept the results if they didn't agree with them. It's become a
religious, not a scientific, argument.