GSIC tests and statistics

[Startz]

In the Challenge section Kramer has forwarded a suggestion for having a number of people try the GSIC test. Basically, the idea is to get 20 people who think they can tell the difference to each give it one try.

My quick check is that with 20 participants, there's about a 1 in 20 chance that more than 70 percent will get it right by pure chance.

So two questions:

(1) If this happened, would it look bad? Assuming there are people who really wan to make JREF look bad.

(2) Can someone check my statistics?

 

[Pyrts]

Eh, I can see a number of ways it could be done, but you'd need double-blinded tests. And I think I'd want more than 20 people.

But there's a much simpler way of doing it -- capture the sound onto a computer and analzye the wave forms. If there's any sort of real difference, you should spot it when you set the two forms on top of each other. You can minimize some of the variables by making the input a simple wire patch into the microphone port of the PC.

Quick, visual, quantitative and not qualitative.

 

[-42-]

But there's a much simpler way of doing it -- capture the sound onto a computer and analzye the wave forms. If there's any sort of real difference, you should spot it when you set the two forms on top of each other. You can minimize some of the variables by making the input a simple wire patch into the microphone port of the PC.

Way too scientific, besides, the magic fairy people simply find something else to do other than bettering the sound quality when a computer is analyzing it.

That's an extremely simple test, if it showed any difference whatsoever well naturally you'd see it plastered all over their [the sellers of the device] website along with possible nobel prizes forthcoming.

 

[Rolfe]

I'm sure I saw posts in the earlier threads saying that someone in the hifi community had in fact done a bit-by-bit comparison of a disc before and after treatment and (surprise, surprise....) the two were identical.

Rolfe.

 

[webfusion]

statistics and lies

If 20 people listen to the GSIC chip, I would assume that all of them will notice no difference, simply because there is no effect to discern -- since the GSIC itself is a fraud and a lie.

Next case.

 

[jmercer]

Re: statistics and lies

If 20 people listen to the GSIC chip, I would assume that all of them will notice no difference, simply because there is no effect to discern -- since the GSIC itself is a fraud and a lie.

Next case.

And if they do notice a difference... it's a conspiracy!

 

[LostAngeles]

Eh, I can see a number of ways it could be done, but you'd need double-blinded tests. And I think I'd want more than 20 people.

But there's a much simpler way of doing it -- capture the sound onto a computer and analzye the wave forms. If there's any sort of real difference, you should spot it when you set the two forms on top of each other. You can minimize some of the variables by making the input a simple wire patch into the microphone port of the PC.

Quick, visual, quantitative and not qualitative.

I believe that when the other people do it, it will be the same baseline test that I'll be performing.

 

[Gr8wight]

In the Challenge section Kramer has forwarded a suggestion for having a number of people try the GSIC test. Basically, the idea is to get 20 people who think they can tell the difference to each give it one try.

My quick check is that with 20 participants, there's about a 1 in 20 chance that more than 70 percent will get it right by pure chance.

So two questions:

(1) If this happened, would it look bad? Assuming there are people who really wan to make JREF look bad.

(2) Can someone check my statistics?

It is important to note that two separate tests are being discussed. The initial baseline test is being conducted in an unblinded setting. The subject, in this case LostAngeles, will listen to an untreated and a treated CD, knowing which is which, in order to determine if she believes she can tell the difference. This is being done because she has never tested the GSIC before, and has no personal opinion regarding its efficacy. If she determines that she can hear a difference between an untreaded and a treated CD, she will proceed to the official preliminary test for the JREF challenge, which involves much more than a simple 50/50 guessing game. Michael Anda had no need of such a baseline test as he has used the GSIC mutiple times, and has clearly stated his belief that he can hear a difference.

In my opinion, having 20 people do an A-B comparison between two CDs will not provide any valuable data, as each person will have a simple 50/50 chance of guessing right, and anyone who has flipped a few coins knows how easy it is to have a run of one result that appears non-random. I would not consider the results significant for a sample size of fewer than 100 respondants.

 

[IXP]

Assuming the 20 trials are independent and properly blinded, If all 20 could pick out the treated disc, I would be very impressed. The probability of this occurring by chance is less than one in a million.

To get the 70% mentioned in the OP (14 correct out of 20), the probablity is 0.058, or about 1 chance in 18.

IXP

 

[LongFuzzy]

This proves that Kramer is working for the GSIC folks, and trying to get more sales!