Ganzfeld million dollar challange?

[Moochie]

The same point as I made before. We seem to be discussing apples and oranges here. The Million Dollar Challenge is not looking to be "taken seriously from a scientific point of view." The Million Dollar Challenge is not a scientific study. The Million Dollar Challenge is not trying to find out whether or not PSI (or any other alleged paranormal ability) exists.

If you would like to see more work done on Ganzfeld type studies, approach a university. The JREF is not interested.

The Million Dollar Challenge is no more and no less than exactly what its name implies. It is a challenge to those who make claims of paranormal abilities to demonstrate those claims. Nothing more. You and Rodney are engaging in one very long Strawman argument. You are criticising the JREF Million Dollar Challenge for not being something it has never claimed to be.

Yes, and it is getting oh so tiresome, like a gnawing toothache that only extraction will ameliorate.

Andy2001 and Rodney, your questions have been asked and answered numerous times. Just what do you hope to achieve by continuing?

If either of you are going to apply to take the MDC, then do it. Your continued quibbling about your pet fantasies is misplaced in this part of the forum.


M.

 

[andy2001]

Here is what I was thinking of:

http://www.internationalskeptics.com/forums/showthread.php?postid=3993129#post3993129
http://www.internationalskeptics.com/forums/showthread.php?postid=3993390#post3993390

It seems to run from one in a hundred to one in a million.

If I were applying, I'd start with one in a hundred.

Linda

Don’t forget Randi needs two tests and the numbers in that link are probably just to pass the first one. If you have two tests at one in a thousand like GzuzKart said you end up with one in a million.


Looking at these tests I get the impression that only the most delusional applicants ever get as far as agreeing a protocol. Some of them are so stupid they seem eager to agree to standards even higher that what Randi requires. It seems only large effect sizes from short tests will do.

 

[andy2001]

Yes, and it is getting oh so tiresome, like a gnawing toothache that only extraction will ameliorate.

Andy2001 and Rodney, your questions have been asked and answered numerous times. Just what do you hope to achieve by continuing?

If either of you are going to apply to take the MDC, then do it. Your continued quibbling about your pet fantasies is misplaced in this part of the forum.


M.

Well some recent posts from GzuzKart and the links from fls do seem to have shed some light on the matter. It seems to win the prize a ball park number is odds of one in a million based on two tests of one in a thousand. It also seems that the effect size will need to be high and based on a short test. I thank those on the tread who helped me reach this conclusion.

 

[Uncayimmy]

The person can have a non random way of picking such as always choosing the first target, or preferring red targets. But if targets are presented randomly the chance score is 25%.

Before the test, yes. After the test and knowing the target results and technique used to guess, no. And we're talking about analyzing the results here, which comes after the test.

Take a very simple test consisting of two coin flips. There are four possible outcomes: HH, HT, TH, TT. We have two people guessing: random guy (RG) and always heads guy (AHG).

Before the test, what are the chances of either guy guessing both results correctly? 25%.

After the test, what are the chances either guy guessed both results correctly? 25%

After the test we look at the results of the two coin flips and see that it came up HH. What are the chances that RG got both right? 25%. What are the chances that AHG got both right? 100%. What if it came up TH, what are the chances that RG got both right? 25%. How about AHG? 0%.

So, let's suppose we use 1,000 coin flips. Getting 700 correct is unlikely (the exact odds don't matter). Before the test RG and AHG have equal chance of getting 700 correct. There's only one way AHG can get 700: we need 700 coin flips to come up heads. RG can get 700 correct with any distribution of heads and tails. Interestingly, the odds are the same for either guy, so if I'm a betting man, I don't care what technique they use.

But as a scientist analyzing the results, I care very much about the distribution of my coin flips and the manner in which my subjects guessed. Suppose both guys got 700. I look at my distribution of coin flips and see that heads came up 700 times, a very unlikely event, but it happened.

If I look at AHG's guesses, it is painfully obvious what he did. I can easily explain why he scored 700. When I look at RG's guesses, I can't explain why he got the score he did, so I may choose to investigate further. If we add a third guesser, always tails guy, we'll know why he scored well below chance.

So, if I'm a Ganzfeld researcher and actually see the percentages I've seen reported here, I'm going to look for all possible causes, not just assume some psi stuff is at work.

 

[andy2001]

Before the test, yes. After the test and knowing the target results and technique used to guess, no. And we're talking about analyzing the results here, which comes after the test.

Take a very simple test consisting of two coin flips. There are four possible outcomes: HH, HT, TH, TT. We have two people guessing: random guy (RG) and always heads guy (AHG).

Before the test, what are the chances of either guy guessing both results correctly? 25%.

After the test, what are the chances either guy guessed both results correctly? 25%

After the test we look at the results of the two coin flips and see that it came up HH. What are the chances that RG got both right? 25%. What are the chances that AHG got both right? 100%. What if it came up TH, what are the chances that RG got both right? 25%. How about AHG? 0%.

So, let's suppose we use 1,000 coin flips. Getting 700 correct is unlikely (the exact odds don't matter). Before the test RG and AHG have equal chance of getting 700 correct. There's only one way AHG can get 700: we need 700 coin flips to come up heads. RG can get 700 correct with any distribution of heads and tails. Interestingly, the odds are the same for either guy, so if I'm a betting man, I don't care what technique they use.

But as a scientist analyzing the results, I care very much about the distribution of my coin flips and the manner in which my subjects guessed. Suppose both guys got 700. I look at my distribution of coin flips and see that heads came up 700 times, a very unlikely event, but it happened.

If I look at AHG's guesses, it is painfully obvious what he did. I can easily explain why he scored 700. When I look at RG's guesses, I can't explain why he got the score he did, so I may choose to investigate further. If we add a third guesser, always tails guy, we'll know why he scored well below chance.

So, if I'm a Ganzfeld researcher and actually see the percentages I've seen reported here, I'm going to look for all possible causes, not just assume some psi stuff is at work.

This is silly. It can not explain for example Dalton 1997 Z score 5.2 using the standard hit miss analysis agreed on before the test. Getting 700 coin tosses from 1000 is a Z score of 12.65. If everyone on the planet tossed the coin 1000 times it unlikely anyone would get 700 correct by chance.

 

[Uncayimmy]

This is silly. It can not explain for example Dalton 1997 Z score 5.2 using the standard hit miss analysis agreed on before the test. Getting 700 coin tosses from 1000 is a Z score of 12.65. If everyone on the planet tossed the coin 1000 times it unlikely anyone would get 700 correct by chance.

What part of "the exact odds don't matter" did you not understand? I stipulated the results as being 700 heads out of 1,000 flips. Therefore, it has already happened. We're only concerned about the results and how our two subjects got there. We're not talking about getting 700 correct answers on a 501 to 499 split in the flips. I'm saying that in order to analyze the data you have to look at the distribution after the fact.

As for explaining Dalton, I'm telling you that you have to look at why somebody seemingly did better than chance. For example, if my test randomly picked C as the target 30% of the time and my subject picked C 100% of the time, does that mean the same thing as both my test and subject picking each target ~25% of the time with my subject being right 30% of the time?

Or what if it turns out that 30% of the time the target happened to be the only picture with a bird in it and my subject always picked a picture with a bird in it. What does that mean? To me it means that I have something else I need to control for in the next test.

 

[andy2001]

What part of "the exact odds don't matter" did you not understand? I stipulated the results as being 700 heads out of 1,000 flips. Therefore, it has already happened. We're only concerned about the results and how our two subjects got there. We're not talking about getting 700 correct answers on a 501 to 499 split in the flips. I'm saying that in order to analyze the data you have to look at the distribution after the fact.

As for explaining Dalton, I'm telling you that you have to look at why somebody seemingly did better than chance. For example, if my test randomly picked C as the target 30% of the time and my subject picked C 100% of the time, does that mean the same thing as both my test and subject picking each target ~25% of the time with my subject being right 30% of the time?

Or what if it turns out that 30% of the time the target happened to be the only picture with a bird in it and my subject always picked a picture with a bird in it. What does that mean? To me it means that I have something else I need to control for in the next test.

If the bird target happened by chance then it will also happen by chance that the bird was not the target and the odds will even out in the long run so this still does not explain the data.

 

[Dancing David]

Fine, but in a controlled test, psi may be weak because there is no strong emotional component. Guessing which of four pictures that the sender was trying to transmit is not the most exciting way for the average person to spend a day. Nonetheless, if the results are in the 30-35% range over thousands of trials in a tightly-controlled experiment, what is the alternative to psi?

Oh, so are you saying that there is no Ganzfeld effect?

Or that there have been thousands of tightly controlled trials?

 

[Dancing David]

This is a ridiculous argument. Nobody is claiming an 80% hit rate for Ganzfeld, and anyone that understands statistics knows that a hit rate only a little bid above chance would be very unlikely to have happened by chance if the sample size is sufficient to give a small enough p value.

Um, not true, there is much more than the p value.

You have to have an effect size that rises above statistical noise. P is sort of related to that but you need a frequency sample of a non-effected group for comparison.

 

[Dancing David]

If the bird target happened by chance then it will also happen by chance that the bird was not the target and the odds will even out in the long run so this still does not explain the data.

Nope you have to show that there was a deliberate control to make sure that there were not multiple cards in a set that had a bird and that the target bird was distributed randomly.

This is a huge problem with the Ganzfeld in general.

In the past they did neither.

 

[Uncayimmy]

If the bird target happened by chance then it will also happen by chance that the bird was not the target and the odds will even out in the long run so this still does not explain the data.

You are not understanding what I am saying. Instead of working really hard to prove me wrong, try working really hard to understand what I am saying.

You are talking about the odds of something happening in advance. Under those conditions we agree. I am talking about what actually happened. This is always known with 100% certainty because it already happened. You're talking about the "long run," which by definition means further trials. I'm talking about the N trials that already occurred. That's a different scenario.

Answer these questions for me:

What are the odds of flipping a coin three times and getting heads three times?

What are the odds of someone randomly guessing three heads?

What are the odds that this person got all three correct?

What are the odds of someone who always guesses heads guessing three heads?

What are the odds that this person got all three correct?

Now supposing that after the fact we look at our coin flips and see that we got three heads.

What are the odds that the random guesser got three correct?

What are the odds that the always head guesser got three correct?

There is no guarantee that any set number of trials will end up with an even distribution of answers. Actually, the odds are that you will not have perfectly distributed answers. If you end up with an uneven distribution, certain patterns of non-random guessing could explain the accuracy after the fact. I am not saying that non-random guessing will change in advance the potential results.

 

[andy2001]

Nope you have to show that there was a deliberate control to make sure that there were not multiple cards in a set that had a bird and that the target bird was distributed randomly.

This is a huge problem with the Ganzfeld in general.

In the past they did neither.

They have computers randomly chose the target. Take 4 photos let the computer randomly pick one as the target. Let the computer randomly pick the order they will be shown. Do all this and it’s random. It’s not huge problem with Ganzfeld in general.

 

[andy2001]

You are not understanding what I am saying. Instead of working really hard to prove me wrong, try working really hard to understand what I am saying.

You are talking about the odds of something happening in advance. Under those conditions we agree. I am talking about what actually happened. This is always known with 100% certainty because it already happened. You're talking about the "long run," which by definition means further trials. I'm talking about the N trials that already occurred. That's a different scenario.

Answer these questions for me:

What are the odds of flipping a coin three times and getting heads three times?

What are the odds of someone randomly guessing three heads?

What are the odds that this person got all three correct?

What are the odds of someone who always guesses heads guessing three heads?

What are the odds that this person got all three correct?

Now supposing that after the fact we look at our coin flips and see that we got three heads.

What are the odds that the random guesser got three correct?

What are the odds that the always head guesser got three correct?

There is no guarantee that any set number of trials will end up with an even distribution of answers. Actually, the odds are that you will not have perfectly distributed answers. If you end up with an uneven distribution, certain patterns of non-random guessing could explain the accuracy after the fact. I am not saying that non-random guessing will change in advance the potential results.

What are the odds of flipping a coin three times and getting heads three times?

7 to 1

What are the odds of someone randomly guessing three heads?

7 to 1

What are the odds that this person got all three correct?

7 to 1

What are the odds of someone who always guesses heads guessing three heads?

7 to 1


What are the odds that this person got all three correct?

7 to 1



Now supposing that after the fact we look at our coin flips and see that we got three heads.

What are the odds that the random guesser got three correct?

7 to 1


What are the odds that the always head guesser got three correct? 1 to 1




There is the answer to you’re silly questions, but they chance nothing. The analysis of Ganzfeld studies may be done after the fact but you would have needed a very improbable series of events for them to have happened with out a real effect. Ideas like someone just happened to get a bird by chance and liked birds can not explain the results. This reasoning is flawed.

 

[fls]

That's what I've always found odd about the assumptions that are sometimes made about the baseline in parapsychology research. Why would anyone assume that people guess in a random fashion?

Linda

 

[andy2001]

That's what I've always found odd about the assumptions that are sometimes made about the baseline in parapsychology research. Why would anyone assume that people guess in a random fashion?

Linda

There no need to make this assumption. If you set up an experiment with odds of 25% because you have 4 photos randomly select one as the target and randomly select the order they will be shown the expected results for choosing the first target or liking red targets or choosing randomly are still 25%.

 

[Uncayimmy]

Now supposing that after the fact we look at our coin flips and see that we got three heads.

What are the odds that the random guesser got three correct?

7 to 1


What are the odds that the always head guesser got three correct? 1 to 1


There is the answer to you’re silly questions, but they chance nothing. The analysis of Ganzfeld studies may be done after the fact but you would have needed a very improbable series of events for them to have happened with out a real effect. Ideas like someone just happened to get a bird by chance and liked birds can not explain the results. This reasoning is flawed.

The questions are not silly. You are making statements without any facts.

In the N Ganzfeld trials, what was the distribution of A, B, C and D for the target?

What was the distribution of A, B, C, and D for the subject selections?

In a control group with no sender, what would be the expected distribution of selections made by human subjects? Hint: It's not equal.

What, if anything, did the target images have in common beyond a perfect distribution?

How about the decoy images?

You are jumping to the conclusion that only psi can be at work without considering any other possible factors. You are arguing in effect, "Psi will result in answers greater than chance. We see results greater than chance. Therefore, psi must be at work."

That's just wrong.

 

[andy2001]

The questions are not silly. You are making statements without any facts.

In the N Ganzfeld trials, what was the distribution of A, B, C and D for the target?

What was the distribution of A, B, C, and D for the subject selections?

In a control group with no sender, what would be the expected distribution of selections made by human subjects? Hint: It's not equal.

What, if anything, did the target images have in common beyond a perfect distribution?

How about the decoy images?

You are jumping to the conclusion that only psi can be at work without considering any other possible factors. You are arguing in effect, "Psi will result in answers greater than chance. We see results greater than chance. Therefore, psi must be at work."

That's just wrong.

Ganzfeld has been subjected to intense peer review to look for flaws. As long as the targets are randomly selected and randomly presented and there is no sensory leakage then there is no other reasonable explanation but psi.

And as for the question about what you call the control group my answer is I would expect it to be biased to whatever was listed as target even if there is no sender. Telepathy is not the only type of psi. If there was no correct choice then I would expect a bias towards the first option shown, but if this is done on a real test it only gets 25% correct.

 

[Uncayimmy]

Ganzfeld has been subjected to intense peer review to look for flaws. As long as the targets are randomly selected and randomly presented and there is no sensory leakage then there is no other reasonable explanation but psi.

You say psi. I say that the statistical difference was because Oprah had precognition and used telekinesis to exercise mind control over the device that selected which images to show.

Please explain to me how either of our theories could be falsified.

BTW, you keep throwing around "random" as if it means "perfectly distributed" when it does not mean that at all. The random chance of being dealt four aces in five card draw is pretty slim, but it happens all the time. If you flip a coin 20 times, only 17% of the time will you get 10 heads. About 50% of the time you will get 9, 10, or 11 heads. I just want to be clear on that.

And as for the question about what you call the control group my answer is I would expect it to be biased to whatever was listed as target even if there is no sender. Telepathy is not the only type of psi. If there was no correct choice then I would expect a bias towards the first option shown, but if this is done on a real test it only gets 25% correct.

In the Ganzfeld studies it is my understanding that the subjects picked the second, third and fourth target presented much more frequently than the first. Was randomization done in the presentation of targets?

Furthermore, why do you insist on repeating what we have already agreed upon? Before the test is run, I agree completely that knowing a guy will always pick heads will not make any difference in his accuracy level.

What I am arguing is how to judge what caused a deviation from strict chance after it happens. You say, "it must be psi" and get butterflies in your tummy because the idea excites you.

I say that one thing to look at is the distribution of the data as it actually happened. You keep arguing that it shouldn't matter of an infinite set of trials. Too bad we're dealing with a finite set, eh?

 

[andy2001]

At point we seem to be heading for an infinite verbal loop. I think I have got as close as possible to getting an answer to my original question. I thank everyone who helped me with that. I will probably not post again on this thread.

 

[jojonete]

So how do you explain the fact that 3145 standard ganzfeld trials conducted during 1974-2004 resulted in 1008 hits (32% hit rate)?

Of all the possible explanations given so far in this thread, my favorite goes along the lines of "failed tests get less publicity, so the test results available for use in a meta-analysis are likely to be the most successful ones".

Now, I hope you'll agree that, until the results of the meta-analysis are replicated in an independent test, they don't carry much weight. So, Ganzfeld proponents should be at this moment working in more tests to replicate the results. As I see it, the point of this whole thread is to ask if those new tests could be valid for the MDC, which of course can't be answered until there's some description of those tests, which in turn is the reason why I keep asking what a Ganzfeld test is like.

The idea is:

• If someone's willing to replicate the results, I'd like to know how the new tests will be conducted, so I can make an idea about how MDC-like they are.
• If nobody's willing to replicate the results, we can assume that Ganzfeld will never be observed again, so it surely won't win the MDC.