Home |
Search |
Today's Posts |
#82
|
|||
|
|||
|
#83
|
|||
|
|||
On 12/19/04 1:50 PM, in article , "Stewart
Pinkerton" wrote: On 19 Dec 2004 15:56:47 GMT, B&D wrote: On 12/18/04 6:32 PM, in article , " wrote: ABX box where I identified absolute polarity 18 times out of 20, which exceeds the 99% confidence level. Isn't 18/20 90%? Still not a bad result, and one that I would start paying attention to absolute polarity, BTW, but what is 99% confidence? No, you need to look up basic statistics. 18 correct out of twenty 'coin tosses' is *way* better than 90% probability of a detectable difference. Ah, misread it - if I select coprrectly 18 times out of 20 times - then I am a 90% accurate detector. |
#84
|
|||
|
|||
"one of the reasons you rejected the results of a blind test I took
in 1984 using an ABX box where I identified absolute polarity 18 times out of 20, which exceeds the 99% confidence level. (You have" In which case you suggest confidence in both the testing method and confidence to have abilities to hear differences during same. Can we then know why you refuse to do such testing on wire, stands, etc. to establish that the subjective enterprise has a reality other then those perceptions originating in the brain alone? You have said you have no reason to be tested but seem to do so because you are satisfied of your nudemonstrated ability to do so and/or for any motivation to do so for publishing reasons. Have you no scientific curiosity, even if you never make the results appear in print? Or is that the reality, you have done such testing and/or done such to others and wish the results not to be known? "Yes, in your _opinion_, Mr. Nousaine, there is no reason for amplifiers to sound different, but as has been pointed out before, your opinion carries no special weight over anyone else's. John Atkinson" Opinion is not relevant when he only referes to an established benchmark of hearing alone testing which shows no reason to reject the conclusion that amps don't sound different, within the parameters often expressed. Views of using the benchmark is opinion, the collective results are not. This is what defines science in part, we don't have to rely on opinion of authorties just because they are seen as such, the research speaks for itsef even if the author of it were not known. Conclusions based on opinion, such as you suggest, are as common and in number to the content of any local bar on any night. Demonstrated evidence used well in support of valid conclusions flowing from it are quite another thing. You have and I have an opinion is ok at the bar, in pursuit of validity in the realy world no such egalatarian ground serves. |
#85
|
|||
|
|||
Hi Ed,
That was the impression I got from those here who have such high esteem for him. I've read all those posts and didn't get any such impression. I don't think that any reasonably rational person who read them without predjudice could get any such impression. Then I got the wrong impression (I have not read ALL the posts). My impression of Randi has been that he is a "debunker of fraud", essentially a "truth-seeker" who aspires to publish the absolute truth. I didn't realize this could be interpeted as name-calling. If that's what you consider to be name-calling, then I admit it I think any reasonably disinterested party familiar with the common falacies of argument would call it that. I guess you're right. I am interested (thus the post), and I am not "familiar with the common falacies of argument". (You must have been an English major. I'm just a firmware engineer - the only arguments I know about are the ones I pass to subroutines ;^) though it appears that you have insulted him more than I ;^) Huh? I wrote, "...Randi, the publisher of the absolute truth...", and you replied, "In fact I've read his stuff for years and it's rather the opposite." My impression was that Randi would not object to my characterization of him, but he would not care for yours. Obvioulsy we tend to have opposite impressions. I encourage Randi in his endeavors; I just think he (and others) could have handled this better. In retrospect, on just about any issue at all, it will virtually unanimously be seen that all sides could have handled things better. Thus, your comment above appears to me to be devoid of content. I bet'cha that's some more of that "common falacies of argument" stuff again. Stop showing off. Sorry if I got your dander up. My "dander" has nothing to do with it. Your name calling did. I'm sure you have chastised those posters who called John Atkinson a "snake oil salesman" just as harshly as you did me for calling Randi "publisher of the absolute truth" (I'm sorry I missed your posts). Regards, Tip |
#86
|
|||
|
|||
"Steven Sullivan" wrote in message
... From my own experiences with inadvertant 'phantom switching' -- where I thought something had changed, with a concomitant 'obvious' change in sound, only to find that nothign had really changed -- I know that having experienced it once doens't mean I never experience it again! Are you certain that the (exact) position of your ears didn't change? |
#87
|
|||
|
|||
On 20 Dec 2004 00:38:46 GMT, "Tip" wrote:
Then I got the wrong impression (I have not read ALL the posts). My impression of Randi has been that he is a "debunker of fraud", essentially a "truth-seeker" Well, that's one aspect of what he does and writes about. who aspires to publish the absolute truth. His position is actually generally in accord with the so-called "scientific method". One important fact about science is that it generally doesn't claim to produce any absolute truths. and I have *never* read any claim by Randi that he is in any way infallible. I am interested (thus the post), and I am not "familiar with the common falacies of argument". Then you are uneducated, no matter what degrees you might hold. (You must have been an English major. I'm a high school graduate of the early 1960's and didn't major in English. I can also read. A good education is available to anyone who has a Public Library in their vicinity. I took advantage of that to, among other things, read up on what does and what does not constitute a good rational argument. I wrote, "...Randi, the publisher of the absolute truth...", and you replied, "In fact I've read his stuff for years and it's rather the opposite." My impression was that Randi would not object to my characterization of him, Actually I am confident that he would be deeply and justifiably insulted by it - as would I if you applied it to me, since it is clearly intended to be ironically insulting. but he would not care for yours. Obvioulsy we tend to have opposite impressions. But *you* admit you have read hardly anything he has written. I, on the other hand have read many articles by him and several of his books. I have, in other words, accumulated some evidence. I'm sure you have chastised those posters who called John Atkinson a "snake oil salesman" just as harshly as you did me for calling Randi "publisher of the absolute truth" (I'm sorry I missed your posts). I didn't criticise you harshly at all, actually. I merely said that you were name calling, which I still believe you were in fact doing. On the other hand it is certainly true that JA makes a living for, among other things, accepting and publishing advertisements for products that I personally think can justifiably be called "snake oil". As does one of our local radio stations around where I live, which accepts advertisements for "magnetic" matresses. Ed |
#88
|
|||
|
|||
B&D wrote:
On 12/19/04 1:50 PM, in article , "Stewart Pinkerton" wrote: On 19 Dec 2004 15:56:47 GMT, B&D wrote: On 12/18/04 6:32 PM, in article , " wrote: ABX box where I identified absolute polarity 18 times out of 20, which exceeds the 99% confidence level. Isn't 18/20 90%? Still not a bad result, and one that I would start paying attention to absolute polarity, BTW, but what is 99% confidence? No, you need to look up basic statistics. 18 correct out of twenty 'coin tosses' is *way* better than 90% probability of a detectable difference. Ah, misread it - if I select coprrectly 18 times out of 20 times - then I am a 90% accurate detector. I don't think it is a case of misreading. And I don't think you understand the principles of statistics yet. If you select correctly 18 times out of 20 times, then there is a very high probability (much more than 90%) that you are not guessing and that there is some real detectable difference. Have they stopped teaching statistics in undergraduate engineering curriculum these days? |
#89
|
|||
|
|||
|
#90
|
|||
|
|||
B&D wrote:
On 12/19/04 12:55 PM, in article , "Ed Seedhouse" wrote: In fact I'd say myself that it's the purveyors of so-called "subjectivism" here, if anyone, who claim to be the "purveyors of absolute truth". They have beliefs about sound quality that, it appears to me, they won't change no matter how much evidence accumulates against them. Part of being a subjectivist is that you can present one with 20 reams of data showing that some piece of gear is sub-par, but if they like how it actually sounds and they think it helps them recreate the live performances they crave to hear it is only so much recycling. I totally agree that it is impossible to design audio components without good objective targets and guidelines as well as good engineering that, again involved measurement and theory. When someone goes to buy the component, however, it is unlikely they will take anything more than their ears and perhaps a tape measure to see if it will fit in their living room space. That's why we say that you could choose your equipment whichever way you like. The ear is a very important instrument - the most important one, actually, but the brain is a rather subjective filter to begin with, too and therein lies the conundrum. There is no conundrum. When it comes to whether the sound is different or not, simply listen with ears only. Note that the believers are saying that the sound is different, but then they fail to back that up with measurements or bias-controlled testing. |
#91
|
|||
|
|||
On 20 Dec 2004 03:55:03 GMT, B&D wrote:
I have always had a question of Randi's "mission" - he is a debunker, but I am not convinced that he really does aspire to publish absolute truth Well that's hardly surprising. Who knows the "absolute" truth? Not I, and not Randi. Is there in fact any such thing as "absolute" truth? I don't think there is, and I believe most scientists don't either. As far as I know science in general makes absolutely *no* claim to discover "absolute" truth. In fact, if Randi were to claim that he is seeking absolute truth that alone would make him very suspect indeed, in my eyes at least. Ed |
#92
|
|||
|
|||
On 20 Dec 2004 00:30:57 GMT, B&D wrote:
On 12/19/04 1:50 PM, in article , "Stewart Pinkerton" wrote: On 19 Dec 2004 15:56:47 GMT, B&D wrote: On 12/18/04 6:32 PM, in article , " wrote: ABX box where I identified absolute polarity 18 times out of 20, which exceeds the 99% confidence level. Isn't 18/20 90%? Still not a bad result, and one that I would start paying attention to absolute polarity, BTW, but what is 99% confidence? No, you need to look up basic statistics. 18 correct out of twenty 'coin tosses' is *way* better than 90% probability of a detectable difference. Ah, misread it - if I select coprrectly 18 times out of 20 times - then I am a 90% accurate detector. No. If you get 18 correct out of 20, there is a better than 99% chance that there was something detectable. 90% would be at about 14 correct out of 20, but it would need only about 60 out of 100 (or thereabouts - sorry, can't do stats in my head any more, but you get the idea - probability increases with the number of trials) You can't do probability calculations by simple division. BTW, note that a 95% probability is exactly that - a probability. If you toss a coin twenty times and get 15 heads, then that will happen on average once every twenty times you try it. So, getting that score in a cable comparison isn't absolute proof, it's just a generally accepted benchmark. Makes me wonder why nobody's tried to win the prize just by sheer guessing! :-) -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#93
|
|||
|
|||
Ed Seedhouse wrote:
On 20 Dec 2004 00:38:46 GMT, "Tip" wrote: I'm sure you have chastised those posters who called John Atkinson a "snake oil salesman" just as harshly as you did me for calling Randi "publisher of the absolute truth" (I'm sorry I missed your posts). ...it is certainly true that JA makes a living for, among other things, accepting and publishing advertisements for products that I personally think can justifiably be called "snake oil". Please note, Mr. Seedhouse, that I have no control over any what ads appear in Stereophile. The "Chinese Wall" that separates a magazine's advertising department from its editorial staff also operates in the opposite direction. I have no knowledge about the ads that appear in an issue of Stereophile until I receive my printed copy. John Atkinson Editor, Stereophile |
#94
|
|||
|
|||
"Stewart Pinkerton" wrote in message
... On 18 Dec 2004 23:32:41 GMT, wrote: But what I fail to understand is how the statistics _know_ what the device under test is? How, when testing loudspeakers for example, a 95% probability of the results not being due to random guessing can be taken as probably meaning there is a real sonic difference, but in the case of amplifiers it cannot. Yes, in your _opinion_, Mr. Nousaine, there is no reason for amplifiers to sound different, but as has been pointed out before, your opinion carries no special weight over anyone else's. I can't speak for Tom, but there has been in existence for about six years, a $5,000 prize awaiting anyone who can differentiate two cables in a level-matched DBT, in any system of their choice and with any music of their choice. The bar has been set at 15 or more correct out of 20 trials, and that has not changed. Interestingly, despite many claims of 'night and day' differences, there have been no contenders. After playing around with my calculator, I notice that you have a 50-50 chance of losing your $5000 after 32 people have taken the test. You should collect $20 for the privilege of taking the test. Even unlikely things happen every so often. Norm Strong |
#95
|
|||
|
|||
Hi Ed,
...I am not "familiar with the common falacies of argument". Then you are uneducated, no matter what degrees you might hold. I thought you had dug yourself into a hole and I was allowing you to get out of it. You don't think anyone actually bought that "common falacies of argument" rhetoric, did you? I wrote, "...Randi, the publisher of the absolute truth...", and you replied, "In fact I've read his stuff for years and it's rather the opposite." My impression was that Randi would not object to my characterization of him, Actually I am confident that he would be deeply and justifiably insulted by it - as would I if you applied it to me, since it is clearly intended to be ironically insulting. Thank you - that solves the mystery. "Truth" is in the eye of the beholder. As you mentioned, the "subjectivists" believe they are telling the "truth". I should have said "facts" instead of "absolute truth" to make my intentions clear. My dictionary defines "an absolute truth" as "not doubted; actual; real" - there was no mention of "irony". However, I appreciate your explanation of how the scientific method does not claim to produce any absolute truths - that's absoutely true! :^) I leave more educated than I arrived. Regards, Tip PS: as a special favor to you, I will never testify that you have ever told the absolute truth ;^) |
#96
|
|||
|
|||
"Please note, Mr. Seedhouse, that I have no control over any
what ads appear in Stereophile. The "Chinese Wall" that separates a magazine's advertising department from its editorial staff also operates in the opposite direction. I have no knowledge about the ads that appear in an issue of Stereophile until I receive my printed copy. John Atkinson" At worst your knowledge of ad content is but one iissue old and a backward perspective to confirm and see trends. While knowing can't be absolute the categories and brands most likely to appear are knowable within a high degree of probability. "Contro" cuts both ways from the editor's chair, content which consistently contridicts marketing claims and manufactured images would be the editorial feedback which also affects potential of reupping on ad space. A tice clock now and again makes little difference and adds a bit of ammo for critics, but wire etc. put in a supportable light would soon sever it's connection with the mag. |
#97
|
|||
|
|||
normanstrong wrote:
"Stewart Pinkerton" wrote in message ... On 18 Dec 2004 23:32:41 GMT, wrote: But what I fail to understand is how the statistics _know_ what the device under test is? How, when testing loudspeakers for example, a 95% probability of the results not being due to random guessing can be taken as probably meaning there is a real sonic difference, but in the case of amplifiers it cannot. Yes, in your _opinion_, Mr. Nousaine, there is no reason for amplifiers to sound different, but as has been pointed out before, your opinion carries no special weight over anyone else's. I can't speak for Tom, but there has been in existence for about six years, a $5,000 prize awaiting anyone who can differentiate two cables in a level-matched DBT, in any system of their choice and with any music of their choice. The bar has been set at 15 or more correct out of 20 trials, and that has not changed. Interestingly, despite many claims of 'night and day' differences, there have been no contenders. After playing around with my calculator, I notice that you have a 50-50 chance of losing your $5000 after 32 people have taken the test. You should collect $20 for the privilege of taking the test. Even unlikely things happen every so often. Norm Strong It seems like the 95% confidence level is established at 15 out of 20 trials. So there is a 5% chance that someone can win $5K by sheer guessing. Should we then set the entry fee to 5% of $5K, which is $250? I'm sure that should not be a deterrent to those who are sure that there is a difference . Or we can lower the entry fee to $50 if the bar is set at 18 out of 20 trials. How did you come up with $20? |
#98
|
|||
|
|||
|
#99
|
|||
|
|||
"chung" wrote in message
... normanstrong wrote: "Stewart Pinkerton" wrote in message ... On 18 Dec 2004 23:32:41 GMT, wrote: But what I fail to understand is how the statistics _know_ what the device under test is? How, when testing loudspeakers for example, a 95% probability of the results not being due to random guessing can be taken as probably meaning there is a real sonic difference, but in the case of amplifiers it cannot. Yes, in your _opinion_, Mr. Nousaine, there is no reason for amplifiers to sound different, but as has been pointed out before, your opinion carries no special weight over anyone else's. I can't speak for Tom, but there has been in existence for about six years, a $5,000 prize awaiting anyone who can differentiate two cables in a level-matched DBT, in any system of their choice and with any music of their choice. The bar has been set at 15 or more correct out of 20 trials, and that has not changed. Interestingly, despite many claims of 'night and day' differences, there have been no contenders. After playing around with my calculator, I notice that you have a 50-50 chance of losing your $5000 after 32 people have taken the test. You should collect $20 for the privilege of taking the test. Even unlikely things happen every so often. Norm Strong It seems like the 95% confidence level is established at 15 out of 20 trials. So there is a 5% chance that someone can win $5K by sheer guessing. Should we then set the entry fee to 5% of $5K, which is $250? I'm sure that should not be a deterrent to those who are sure that there is a difference . Or we can lower the entry fee to $50 if the bar is set at 18 out of 20 trials. How did you come up with $20? The probability of 15 correct out of 20 is 2.1%. IOW, the chances of not doing that well are 97.9%. The p of not doing that well 32 times is thus 0.979^32 = .507. So how did I come up with $20. Simple. I dropped a zero in my haste to post. :-) Sorry, Norm |
#100
|
|||
|
|||
In article , Stewart Pinkerton
wrote: [from an earlier post in this thread] No, you need to look up basic statistics. 18 correct out of twenty 'coin tosses' is *way* better than 90% probability of a detectable difference. [and now in a more recent one] No. If you get 18 correct out of 20, there is a better than 99% chance that there was something detectable. Actually Mr. Pinkerton needs to look up basic statistics. Here he makes the same error that Mr. Atkinson and Mr. Krueger frequently do. The claim that getting 18 correct out of 20 means there is a better than 99% chance that there was something detectable is essentially: (1) IF someone gets 18 correct in 20 trials THEN P[ there is something detectable ] .99 . What is true is that: (2) IF someone is guessing (or flipping a fair coin) THEN P[ at least 18 correct in 20 trials ] is about 0.000201, and P[ exactly 18 correct ] is about 0.000181. (1) and (2) are NOT equivalent yet they are often confused, especially by people with no mathematical and statistical training. BTW unless you are a Bayesian you probably wouldn't even write something like (1), as it is meaningless to a classical (frequentist) statistician. For a frequentist the probability that someone was guessing is either 1 or 0, but he just dosn't know which of those it is. [There are subjectivist/objectivist issues outside of audio. ;-) ] If you toss a coin twenty times and get 15 heads, then that will happen on average once every twenty times you try it. If Mr. Pinkerton wants to bet that on average he'll get 15 heads in twenty tosses once every twenty times he tries it, he may find lots of people willing to bet against him. The actual probability of exactly 15 heads in 20 tosses is .0148 (about 1/68) and the probability of at least 15 heads in 20 trials is 0.0207, or 1/48. Maybe Mr. Pinkerton should learn some statistics. ;-) |
#101
|
|||
|
|||
In article , "normanstrong"
wrote: "chung" wrote in message ... normanstrong wrote: After playing around with my calculator, I notice that you have a 50-50 chance of losing your $5000 after 32 people have taken the test. Close, but you have to watch round-off error. x P[at least 1 winner in x runs of 20 trials] -------------------------------------- 32 .4879 33 .4985 34 .5088 It seems like the 95% confidence level is established at 15 out of 20 trials. So there is a 5% chance that someone can win $5K by sheer guessing. Should we then set the entry fee to 5% of $5K, which is $250? But the probability of winning is not 5%. The binomial distribution is discrete, so its natural significance tests come only in certain sizes, e.g., 14/20 gives .0577 and 15/20 gives .0207 (actually .02069473). Using 15/20 we get an expected value of (.02069473)(5000) + (.97930527)(0) = $103.47 as an entry fee for a fair game. BTW, there is a difference between a *significance level* of 5% and a *confidence level* of 95%. Or we can lower the entry fee to $50 if the bar is set at 18 out of 20 trials. P[ 18 or more correct in 20 trials ] = .0002012253 so expectation for fair game is (.0002012253)(5000) = $1.006126 How did you come up with $20? The probability of 15 correct out of 20 is 2.1%. IOW, the chances of not doing that well are 97.9%. The p of not doing that well 32 times is thus 0.979^32 = .507. So how did I come up with $20. Simple. I dropped a zero in my haste to post. :-) ???? |
#102
|
|||
|
|||
In article , Chung wrote:
I don't think it is a case of misreading. And I don't think you understand the principles of statistics yet. If you select correctly 18 times out of 20 times, then there is a very high probability (much more than 90%) that you are not guessing and that there is some real detectable difference. Have they stopped teaching statistics in undergraduate engineering curriculum these days? If that's where people pick up the sort of misunderstandings that some posters have of statistics, then maybe they should stop teaching statistics in undergraduate engineering curricula. ;-) Perhaps you could explain exactly how you get a numerical value (e.g., much more than 90%) based solely on selecting correctly 18 times out of 20 times, for the probability that you are not guessing and that there is some real detectable difference? |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Randi look at Stereophile message exchange | High End Audio | |||
Stereophile Tries To Come Clean About The DiAural Fiasco | Audio Opinions | |||
Does anyone know of this challenge? | High End Audio | |||
Note to the Idiot | Audio Opinions |