Re: Re: Re: Which way is up??
Ah, but if your results are truly random these things will occur every now and again by chance, the only way to eliminate this is the use of more tests on more people.
The way to think of this is to ask the question "If you have already tossed a coin 9 times and got 9 heads what is the chance of tossing it again and getting a head?" The answer is actually 50%, but most people will say it is less than this because they intrinsically believe that if they have thrown that many heads the chances of getting another decrease, because they expect the results to become normalised. Yet another way perceptions effect expectations.
Rob M said:
Ah, but if your results are truly random these things will occur every now and again by chance, the only way to eliminate this is the use of more tests on more people.
The way to think of this is to ask the question "If you have already tossed a coin 9 times and got 9 heads what is the chance of tossing it again and getting a head?" The answer is actually 50%, but most people will say it is less than this because they intrinsically believe that if they have thrown that many heads the chances of getting another decrease, because they expect the results to become normalised. Yet another way perceptions effect expectations.
Stats 104 .... Independence
Damn, damn, damn you pinkmouse !
I said back on page "something" that this was actually quite difficult, and hinted at "things I wasn't going to go into".
Well, the above is one of those things.
Tossing a coin is a truly independent event. Yes, the 10th toss has equal probability of being H/T = 50%
Question. If GoldenEar has guessed 'A' nine times, on the 10th choice do you believe the probability of him/her choosing 'A/B' remains independent at 50%?
If you think the answer is YES ...... I have some very bad news for you.
Bummer .... eh!
cheers
mark
Damn, damn, damn you pinkmouse !
I said back on page "something" that this was actually quite difficult, and hinted at "things I wasn't going to go into".
Well, the above is one of those things.
Tossing a coin is a truly independent event. Yes, the 10th toss has equal probability of being H/T = 50%
Question. If GoldenEar has guessed 'A' nine times, on the 10th choice do you believe the probability of him/her choosing 'A/B' remains independent at 50%?
If you think the answer is YES ...... I have some very bad news for you.
Bummer .... eh!
cheers
mark
There has been a lot of disparaging talk on this thread and nw_avphile has been taking quite a bit fire. I am quite sympathetic to the objectivity of his purpose. There are many passive parts that I have changed in certain systems that I can attest had absolutely no audable effect on sound (probably because of the aforementioned masking effects of the other components) but those very same component exchanges in other systems have definitley been notable. The black gate capacitor is a typical favorite to hold as an example of a contraversial component. I have definitely heard a difference with this cap when it replaced others but was it always an improvement? Was it worth the substantial price premium? These kind of questions are of a very personal nature and I think the testing proposed by nw_avphile makes eminent sense and I believe a great many of the membership on this thread should endeavor to prescribe to his advice. To me, all he is saying is listen before you take a leap of faith on the opinion of some alledged authority. I am not so ready to wholesale dismiss all premium components as snakeoil but I think it a good idea to be skeptical of all the claims. They may be true for some but not for all. Just like some people can see a stereographic image in those "magic eye" prints and some can't, some people can hear and place stereographic sound and their nuances where some can't. The only way to have an honest assessment of whether what we hear is not a product of our expectations is with these objective tests. A good hearing test before you do wouldn't be such a bad idea either. It might help explain some of your findings.
Fred Dieckmann said:
Yes, that matters too, but I am inclined to think that the actual
recording procedures used in later recordings are often an
even greater reason for bad overall sound. Using myriads of
close-up micophones and then mixing this together to a
stereo image is often an obvious problem in many recordings,
Deutsche Grammophon being almost consistently bad from
the 70s (although, even they have exceptions). Of course,
these more compex recording procedures required more
complex electronics, making op-amps were tempting for many
designers. I would, by the way, guess that most or all studio
equipment were still valve-based in the late 50s and early 60s,
so some people might be tempted to substitute semiconductors
for op amps in your text.
Re: Stats 104 .... Independence
Be gentle with me, my formal stats training is about 15 years old!
Well, that depends on the test
If GoldenEars is truly guessing then of course there is a 50% chance...
If there is any type of experimental procedural error, or some difference is being noted, then of course not
mefinnis said:
Be gentle with me, my formal stats training is about 15 years old!
Well, that depends on the test
If GoldenEars is truly guessing then of course there is a 50% chance...If there is any type of experimental procedural error, or some difference is being noted, then of course not
Der Tonmeister.
Hi,
That, plus the fact that in those days a record was the engineers' best shot at duplicating an actual concert or performance.
Not just for classical music but even today I often find jazz recordings much better done than what we'd expect, same for blues.
These are often none commercial recordings done by engineers that actually care about music...
Here you can find an analogy with some, not all makers of high-end audio, they do care...
Re: Euphonic distortion...come on guys do you actually know what you're on about?
If distortion is present than its' harmonic content, degration is important...even harmonic, odd harmonic and such.
I'd rather listen to an amp with 2% distortion mainly composed of even harmonics than an amp with the same % of distortion but mainly composed of odd order harmonics.
As for feedback, well yes most of the time you can't do without to please the measurement brigade and given a choice I'd use local feedback to linearise the various stages and if must be global feedback to lower the Zo of the design...
Feedback is a nice " under the rug " technique but surely not a cure all.Lest yo forgot.
Ciao,
Hi,
That, plus the fact that in those days a record was the engineers' best shot at duplicating an actual concert or performance.
Not just for classical music but even today I often find jazz recordings much better done than what we'd expect, same for blues.
These are often none commercial recordings done by engineers that actually care about music...
Here you can find an analogy with some, not all makers of high-end audio, they do care...
Re: Euphonic distortion...come on guys do you actually know what you're on about?
If distortion is present than its' harmonic content, degration is important...even harmonic, odd harmonic and such.
I'd rather listen to an amp with 2% distortion mainly composed of even harmonics than an amp with the same % of distortion but mainly composed of odd order harmonics.
As for feedback, well yes most of the time you can't do without to please the measurement brigade and given a choice I'd use local feedback to linearise the various stages and if must be global feedback to lower the Zo of the design...
Feedback is a nice " under the rug " technique but surely not a cure all.Lest yo forgot.
Ciao,
Re: Re: Stats 104 .... Independence
I thought that was against forum rules
You may have missed the subtle point in my post. IF GoldenEar is guessing he will be influenced by his previous choices.
Intuitively/subconsciously we all know it is unlikely in an A/B comparison that the test is likely to involve all 'A' or all 'B'. So, if we are guessing and we have choosen a lot of 'A's, then the probability we will choose 'B' becomes > 50%
In summary, the subsequent event is dependent upon the previous event and the binomial distribution I gave you above does not become a true representation of the likelihood of a given set of observations.
This type of statistical analysis can be done but I trust you see it is more difficult than first thought.
regards
mark
pinkmouse said:
I thought that was against forum rules
Well, that depends on the test
You may have missed the subtle point in my post. IF GoldenEar is guessing he will be influenced by his previous choices.
Intuitively/subconsciously we all know it is unlikely in an A/B comparison that the test is likely to involve all 'A' or all 'B'. So, if we are guessing and we have choosen a lot of 'A's, then the probability we will choose 'B' becomes > 50%
In summary, the subsequent event is dependent upon the previous event and the binomial distribution I gave you above does not become a true representation of the likelihood of a given set of observations.
This type of statistical analysis can be done but I trust you see it is more difficult than first thought.
regards
mark
Brett said:
...i have had the oportunity to examine cheevers thesis, (thanks Brett), and i fear the initial assumptions made, and references cited to justify the work are fataly flawed.
He gives prominance to a subjective review of the cary amp. in 'stereophile', whose lack of intellectual rigor and consistency is legend....for instance, the cary amps. feedback is given fulsome credit for the amps. performance, but we need only recall the Halcro dubbed the best amp. in the world by the same periodical, uses a hefty feedback factor.
Mr. Cheever cites Matti Otala's work on TIM, but sadly appears to
not to be cognisant of the fact that TIM, SID have been demonstrated to be a red herring of the Monster variety.....see P. Baxandal series in wireless world 1978-1979, whose name is conspicuosly missing from this thesis....
...considering that the author subsequently cites Dr. Cabot of Audio Precision at a later stage, i find it inexplicable that he failed to mention an extremely important AES paper by the Dr. 'Comparison of non-linear distortion measurement method' ,which is freely available for download, subject to registration at:
http://audioprecision.com/publications/technical_papers/index2.htm
I have difficulty in accepting that this ommision was accidental, as many figures in this thesis are infact directly copied from other publications on the audio precision website.....
Re: Re: Re: Stats 104 .... Independence
Ah, but if it is a true double blind test, he will not know the results of his previous answers, only if the methodology of the experiment is wrong will he have any information that will bias his next answer. There should be no dependance on previous results.
The fact that one has chosen more As or Bs will not matter if the random selection of the components under test comes up as all As or Bs, which as we know from the coin experiment can happen
mefinnis said:
I thought that was against forum rules
Ah, but if it is a true double blind test, he will not know the results of his previous answers, only if the methodology of the experiment is wrong will he have any information that will bias his next answer. There should be no dependance on previous results.
The fact that one has chosen more As or Bs will not matter if the random selection of the components under test comes up as all As or Bs, which as we know from the coin experiment can happen
Thanks for the support nania... I'm futher encouraging folks here to try some **blind** listening to have more perspective on how psychological factors can otherwise bias what they hear. Once they do, I suspect it will change the way they listen (when doing comparisons) in the future.nania said:
While we're hashing out the details of proper blind testing, it's still a very valid way to demonstrate if differences likely exisit.
Wasn't Mr Otala under contract (a consultant) with Harmon Kardon when he published that paper? Considering HK waved it around as The Next Big Thing in amplifier design in all their marketing materials one has to wonder about his motives for writing it when his paycheck was coming from HK--an amplifier manufacture?mikek said:
Baxandal also pioneered the null difference test just before that period--a test that's very revealing of TIM, etc. under real-world conditions.
The Null against the blind test...
This advice is good for all things in life...try to have "your" opinion about the things!!
Anyone will marrie a girl because a friend tell that she is a good lover!!!
For me the blind test is not so valid as the null test...because in the blind test we are limited by the resolution of the loudspekers...they can turn similar the sound of two diferents amplifiers...
We can never forget that in a blind test we are listening to a complete system...and if specialy the speakers haven't enough resolution...the diferences fade way!!
So for me...null is king!!!!
This advice is good for all things in life...try to have "your" opinion about the things!!
Anyone will marrie a girl because a friend tell that she is a good lover!!!
For me the blind test is not so valid as the null test...because in the blind test we are limited by the resolution of the loudspekers...they can turn similar the sound of two diferents amplifiers...
We can never forget that in a blind test we are listening to a complete system...and if specialy the speakers haven't enough resolution...the diferences fade way!!
So for me...null is king!!!!
Re: Independence
I will grant you this, however, sit back and think about how you are going to set this comparison up. What "question", exactly, are you going to have the listener answer?
My point was you may easily find yourself in a situation where "guesses" are not truly independent.
Seemingly simple things like this can have a significant influence on how the results have to be analysed. Which might explain why it is a branch of science in it's own right
mark
I will grant you this, however, sit back and think about how you are going to set this comparison up. What "question", exactly, are you going to have the listener answer?
My point was you may easily find yourself in a situation where "guesses" are not truly independent.
Seemingly simple things like this can have a significant influence on how the results have to be analysed. Which might explain why it is a branch of science in it's own right
mark
I've only been gone for a day, but my, oh my, how prolific you guys have been!
I've been asking that question for over 10 years now, and the only answer I get is the sound of crickets chirping.
For the fellow a few pages back who brought up wine-tasting (a subject near and dear to my mortgage payments these days), rapid comparisons are, in fact, the ones that give tasters the best acuity. Same with every other bit of sensory research of which I'm aware.
I've been asking that question for over 10 years now, and the only answer I get is the sound of crickets chirping.
For the fellow a few pages back who brought up wine-tasting (a subject near and dear to my mortgage payments these days), rapid comparisons are, in fact, the ones that give tasters the best acuity. Same with every other bit of sensory research of which I'm aware.
- Status
- Not open for further replies.