Re: Ludwig: Commentary on Mc book by Kessler


"Clyde Slick" <Mr.clydeslick@xxxxxxxxx> wrote in message
news:4a8b4155-9953-469c-8dc4-7b34790ce5ce@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
On 3 Iul, 19:56, "ScottW" <Scott...@xxxxxxxxxxx> wrote:
"John Atkinson" <stereoedi...@xxxxxxxxxxxxx> wrote in message


Again, ScottW, we have an instance where I have
practical knowledge of something whereas you only
know what you have read on the Internets but somehow
still feel your opinion has weight. First, there was the
topic of blind tests, which you have never taken part in
but I have;

and I posted my issues with your test and you
neglected to reply. Funny how that works.
But, FWIW, the statistics in blind tests are very
similar to reliability testing, something I am very
experienced with.


: No they are not.

Lol, I don't recall you mentioning your experience
in reliability engineering Art.
This comment reveals that it must indeed be limited
as they are indeed similar.

:The mechanics and methodology
: of the statistics are similar,

Which is exactly what I said.

: but the purpose for which the
:staistics is used is quite different,, the meaning of the results
:of those statistics is different, and the use to
:which they are put is much different.

:For example, in doing DBT for the purpose of
:your own consumer selection, it is absolutely ludicrous
:and nonsensical to use a standard of 95 % confidence

So maybe you'd like to tell what confidence most
component reliability testing results are projected to?

:Your consumer choice should be based upon the
:predonderance of the evidence. Say, you do a test on
:a piece you ahve and a piece you are thinking of getting,
:and the result is with 80% certainty that
:you can tell a difference

80% certainty? Odd combination of words not
often heard in statistics IME.

: say you take that result and decide NOT to investigate
:the new piece for possible qualitative imporovements through
:comparative listening, because you didn't meet the 95% threshhold.

I'd suggest you run a few more trials to improve your confidence
if you felt that 80% confidence was insufficient to act upon.
What would you suggest Art?
I'd also suggest that if you ran a 1000 trials and got just enough
right to establish with 95% confidence your results were not due to
chance, the difference is still pretty small and obviously not detectable in
all circumstances.

:What you have done is eliminate consideration'of a piece where you
:were
:80% confident you could here a difference.
:That is, you were shown that it is much more likely than not that you
:heard a difference.

Setting a confidence limit for a trial to terminate is trivial.
You can set a limit for which a conclusion is reached
(positive or negative) and set a limit for which additional trials
are necessary.
This argument you pose is a silly strawman.
For example. If we planned to test 100 parts for 1000 hours
under some accelerated conditions and 5 parts failed in the first
10 minutes, we would terminate the test. The outcome is established
without needing 6 additional weeks.

:Yet, by blind adherence to scientific standards,

What standard is this? Please provide a reference.

:you blundered
:into stupidly eliminating consideration of something that was much:
:more likely
:than not different sounding than your current piece.

:The point is, you need to have some common sense and perspective
:rearding you ultimate purpose

Of course you do. and the idea you raise that some preestablished
limit determination nullifies the entire concept is AFAIAC,
quite silly.
If you don't like 95% confidence, use 80% for your purpose.
Still when the results are only 75% conifident they are not due to
chance, what will you do?

:for these investigations., and even if you choose
:to proceed with these methods, use them
:in harmony with your goals and purposes.
:You are NOT sending a man to the moon.
:You are only selecting the best sounding stereo.
:Don't reject something jut because you are less
:than 95% sure.

I am not bound to 95% confidence.
Though if the trend of positive results continues,
you will get there with enough trials.
I don't consider the choice of confidence limits a big
deal. Depends on your purpose. If you intend to
convince others and they require 95% to be
convinced, then you have no choice for that purpose.



:And then there is this:

People can claim many things, but bowing to silly claims
and threats of lawsuits as the just and proper thing to do
as you apparently claim, is nonsense.



:THANK YOU JESUS!!!!!!!!

:you seem to be against the use of lawsuits as 'threats'.
:since it is nonsense to bow to them.

I am against the threat of a lawsuit, yes.
John's argument is indeed, quite strange.

ScottW


.

Relevant Pages
Loading