wrote in message ...
Harry Lavo wrote:
wrote in message
...
Harry Lavo wrote:
My test is a standard research test, used broadly all over the world
in
many
fields.

But never to test for *difference*. It would be an absolutely terrible
test for difference. And just because one researcher used it one time
for that purpose and got the result he (and you) wanted doesn't change
that.


Absolutely it is used to test difference. If it doesn't measure
statistically significant, the difference hypothesis is not supported.
Try
telling a drug company it doesn't measure difference from a placebo!

Drug companies aren't measuring difference. They're measuring
effectiveness. Does this antibiotic cure infections? Does this cancer
drug reduce the size of tumors? Measuring difference would be: Does
this drug have any physiological effect on the body whatsoever?


And how do they measure effectiveness. By the *difference* in physical
phenomenon, and by the *difference* in self-reported behavior, that's how.
When the drug company says this drug reduces the risk of heart attack by
25%, it has found a statistically significant difference between the samples
of a magnitude of at least 25%. Likewise, if 200 people rate amp "A" as a
4.0 on a five-point scale with regard to naturalness in reproducing violins,
and another 200 people rate amp "B" as a 4.3 on this same scale, and the
statistical test for difference indicates that this is statistically
significant at some level (usually 95%), then it can be said that Amp "B"
has a more natural violin sound than Amp "A". The statistical test used is
specific to the scalar technology used.

There is no real difference between the drug comapny tests and the musical
tests, except that the drug tests have objective incidences to meaure (as
well, often, as subjective) whereas the music reproduction test is clearly
all subjective. But that is a result of the fact that music itself is
subjective, and *cannot* be measured objectively. The closest you can come
perhaps is to substitute some kind of psychophysiological measurements.


And, no, you cannot use a drug test as a difference test, because you
cannot presume that a drug that fails to cure an infection therefore
had no physiological effect.


Sure you can. If the drug and the placebo (or the control drug) yield the
same incidence of effective cure, there is no difference. The drug doesn't
work (placebo) or at least work any better (control drug). Simple as that.


BTW, now you're talking about monadic tests. Previously, you've touted
what you called "proto-monadic" tests, a la Oohashi. Which is it,
Harry? You say we need a reference against which to compare ABX's
results, but you can't even agree within your own mind about what that
test should be.

Monadic tests are the gold standard because their is no test order bias, but
they require the largest sample size and are those fairly impractical except
for deep-pocketed, major studies. Proto monadic *is* a monadic test, with a
*comparative* tagged directly on the end. Thus one can often get by with
smaller sample sizes because there is yet a second measure of difference.
At least some researchers favor it for this reason under some circumstances.
On the other hand, proto-monadic testing has a strong order bias that has
to be controlled. It's a judgement call, but either are preferable to
quick-switch, comparative testing IMO because they intrude less into normal
listening patterns. Both rely on after-the-fact recall and rating.

One of its strengths is that it goes beyond difference to get at where and
how and why there is a difference. But only if there is a statistical
difference.

The ABX is an audio-specific test. Moreover, it is used, as the
information Harmon Kardon's current use of it.

Not for difference.


Of course for difference.

Oh, this is rich. Now you're touting Harman as the model of your
monadic ideal. Harman doesn't use monadic tests. They use
QUICK-SWITCHING preference tests (not difference tests!). That's why
they built their listening lab, Harry--so they could do quick-switching
tests with speakers. I presume you're now going to tell us that their
findings aren't validated.


Did you miss John Atkinson's recent posts that upon his recent visit to
Harmon's research facitility, they had switched their speaker testing to
monadic, evaluative testing using rating scales? I'm not holding them up
as a model for anything, simply commenting that they are at least one
company using monadic testing in the audio field. Which refutes your claim
that none are.


They are profiling speakers against the profile
of know reference speakers. If there is no statistical difference, any
differences in profile are useless (in a scientific sense). Of course,
the
Harmon people may set a lower statistical standard for such testing
because
they find even directional information helpful. But that is always a
standard development dilemma.

It is much less
controversial test on the face of it, since it minimizes disruption of
normal listening patterns.

ABX is "controversial" the way evolution is controversial. There is
science, and then there are peole who wish that science were not so.


Once again, name calling, however subtle in nature. I am not
anti-science.
I am against bad science parading as good science.

Hey, you're the one who started with the religious insults. And anyone
who professes the authority to "validate" science by ignoring the parts
he just doesn't like is anti-science, in my book.

Religous insults? You mean when I point out that an unwillingness to
consider the underlying premises of the ABX test turns promotion of such a
test (without validation) into a profession akin to religion? As opposed to
true science? If that's your claim, I stand convicted.


The only thing one might fault it for is
sensitivity, but with this type of test sensitivity is simply a matter
of
numbers...need more sensitivity, add more people.

As a test for difference, more numbers wouldn't solve the sensitivity
problem. You're talking about a test that couldn't determine that LP
and CD are sonically different! And you think such a test should be the
gold standard for judging the accuracy of all other difference tests?

Believe me Bob, it tests difference. But it tests subjective musical
impression difference, not sound pressure levels. Can *you* see a
difference?

Yeah, and that's why ABX is better. ABX doesn't JUST test for "musical
impression," it tests for any difference at all. And you can't identify
a single audible sonic difference that ABX tests can't distinguish.

How would you know that? A proper control test has never been done.


As for your own "proposed" test, which is it:
1) monadic, like the pharmaceutical companies?
2) proto-monadic, like Oohashi?
3) quick-switching preference, like Harman?

Or does it depend?

I can't answer because your list is erroneous. I've answered the 1) vs. 2)
above. And 3) simply indicates you missed out on, failed to notice, or for
some reason dismissed John's reporting.


Actually, it doesn't. None of those three tests could be used to
confirm an audible difference between LP and CD. An ABX test could do
so easily. So much for validation.

We aren't looking to determine differences, Bob. We're looking to evaluate
audio components sonic signatures and subjective shading of musical
reproduction. And there has been no confimation that ABX or a straight AB
difference test can show up all the various shadings that show up in
longer-term listening evaluations.