twisted statistics?

[Guest]

Is it true the claim of those who made the clairvoyance test at Princeton University (supposedly repeated at Friburgo and Duke Universities), that as they took a BIG set of tries , ("millions of guesses") the expected distribution of hits of those trying to guess the outcome of the next virtual throwing (coins, I asume) should be exactly 50% if there is not a real psi power?
As thousand of applicants had produced 50,2% to 50,5% of hits, that is THE proof that some clairvoyance exists in every human.

From my statistics-ignorant position I see some problems here. First, did really each applicant guessed "millions" of hits each one?. That should be the condition necesary to validate this "big samples aproximate distribution to 50%" claim.
Am I wrong?
Second, is it true that 0,2 to 0,5% away from 50% is statistically significative?

 

[JamesM]

Second, is it true that 0,2 to 0,5% away from 50% is statistically significative?

It could be, if the size of the sample is large enough.

 

[T'ai Chi]

Is it true the claim of those who made the clairvoyance test at Princeton University (supposedly repeated at Friburgo and Duke Universities), that as they took a BIG set of tries , ("millions of guesses") the expected distribution of hits of those trying to guess the outcome of the next virtual throwing (coins, I asume) should be exactly 50% if there is not a real psi power?

You might want to read their papers to get their exact claims.

Wouldn't you expect that for a larger number of trials, that the true percentage would approach 50%? For millions of trials, one would expect the percentage to be essentially 50%, not a little more than 50%.

 

[Eos of the Eons]

So what exactly is clairvoyance anyway. They say up to 25% of the population is more sensitive to their surroundings. Stimuli affects them more, and they notice more about everything around them.

I can tell it's going to rain just by how I feel sometimes, but then I go outside and understand why I feel that way. It's perception more than psychic abilities.

Heck, even some people taste more in their food than others. What is super sweet to one person is just barely sweet to another.

'psychic' abilities aren't anything but how a person is built to perceive the world around them. People who say they can talk to the dead are scam artists. The sixth sense is bunk. So I can feel a change in whatever pressure when it's going to rain...but even a barometer can do that.

 

[Mercutio]

Re: Re: twisted statistics?

Wouldn't you expect that for a larger number of trials, that the true percentage would approach 50%? For millions of trials, one would expect the percentage to be essentially 50%, not a little more than 50%. [/B]

Depends. Are the trials truly independent? If the millions of trials are actually built up of thousands of sets of hundreds of trials, some small biases could conceivably be built in, if the researchers are not careful. As one example (not saying it happened, but it is also not the only way a bias could enter), if subjects are allowed to choose when to end a session, and they like to end on a success rather than a failure, the very fact that each session ended with a success (or, perhaps, the first failure after a string of successes--again, there are many possibilities, and we would need to see the precise methodology), that small bias, which would be undetectable when comparing to a sampling distribution based on hundreds of trials, would be significant when compared to a sampling distribution based on millions.

So, yeah, for millions of trials we might expect essentially 50%, but we did not have millions of independent trials. The question is, does the difference make a difference? An innocent quirk of running the sessions could be enough to cause a significant effect.

 

[rockoon]

Is it true the claim of those who made the clairvoyance test at Princeton University (supposedly repeated at Friburgo and Duke Universities), that as they took a BIG set of tries , ("millions of guesses") the expected distribution of hits of those trying to guess the outcome of the next virtual throwing (coins, I asume) should be exactly 50% if there is not a real psi power?
As thousand of applicants had produced 50,2% to 50,5% of hits, that is THE proof that some clairvoyance exists in every human.

From my statistics-ignorant position I see some problems here. First, did really each applicant guessed "millions" of hits each one?. That should be the condition necesary to validate this "big samples aproximate distribution to 50%" claim.
Am I wrong?
Second, is it true that 0,2 to 0,5% away from 50% is statistically significative?

The chances of a single guesser getting exactly 50% right in 1 million trials is extremely small. If there is no such thing as psychic voodoo guessing skills then the chance that he will be right MORE THAN 50% of the time is almost precisely 50%

This can be reduced to a simple coin flipping model. If 20 people each make 1 million guesses and these people are compared, the odds are that half of them were right more than 50% and half were right less than 50%. But the sample on THIS data is only 20 strong, not 1 million strong. Ie, it has a high standard deviation.

20 participants all gettings > 50% accuracy is about a million to one shot (2^20-1:1 against).

If only 60% of the 20 people scored > 50%, thats like a 2:1 shot or something. I'm not going to do the precise math.. its clearly not even close to abnormal if only 60% beat the coin.

 

[Guest]

Thats the point.
As I (try to) understand it, to be 50% exactly, it would be necesary to every applicant to make millions of guesses. That's not the same to take 10.000 applicants and get them to make 100 guesses, each. Right?

 

[slimshady2357]

I have a question.

Say they did billions of trials, or however are necesary such that 50.3 % correct would start to become statistically significant.

If one were to say that 50.5% correct proved there was psi, what would 49.5% correct prove?

That people have anti-psychic ability?

Would the psi people suddenly jump up and accept that psi is proven not to exist if the numbers were equally as far from the expected result, but in the negative?

Adam

 

[JamesM]

If one were to say that 50.5% correct proved there was psi, what would 49.5% correct prove?

That people have anti-psychic ability?

I suspect that some would quite possibly interpret it in that way. 'Negative psi' effects are referred to in parapsychological circles.

 

[ingoa]

Actually it would be a miracle if exactly 50% would be reached. There will be certain spread of results, probably Gaussian (depending on the actual test itself).

What statistics give you is either a probability of
- reject ESP, but ESP is still true
- accept ESP, but ESP is still false

But even that does't help you very much. One test (even with millions of coin tosses [btw, has somebody calculated how long that takes? ] will give you one data point in the end. Might be a good start to ask for more funding, but it will not prove anything. Only repetition with similar results will give in the end the (hopefully) correct answer.

 

[Number Six]

Any difference is "statistically significant" if the sample size (in this case, the number of trials) is large enough. Efforts are made in studies to reduce biases but in general small biases aren't a big deal and extremely small biases aren't even considered. In a study with millions of trials extremely small biases completely would ruin the study.

 

[rockoon]

Thats the point.
As I (try to) understand it, to be 50% exactly, it would be necesary to every applicant to make millions of guesses. That's not the same to take 10.000 applicants and get them to make 100 guesses, each. Right?

Millions? There is no amount of guesses you can pick beforehand to ensure exactly 50% right. And certainly if you pick an odd number of guesses, its absolutely impossible.

 

[T'ai Chi]

Re: Re: Re: twisted statistics?

Depends. Are the trials truly independent? If the millions of trials are actually built up of thousands of sets of hundreds of trials, some small biases could conceivably be built in, if the researchers are not careful.

An innocent quirk of running the sessions could be enough to cause a significant effect.

Great points Mercutio. I think that the experimental design has to be very sound.

Honorton did a study showing that there is a significant relationship between 'psi' experiment quality over the years.

 

[T'ai Chi]

If one were to say that 50.5% correct proved there was psi, what would 49.5% correct prove?

That people have anti-psychic ability?

That is a very good question.

Personally, I'd say that before the experiment, the experimentee (the one being tested) has to state whether he/she is going for low or for high output of 1's from a RNG (for example).

 

[Dancing David]

Re: Re: twisted statistics?

You might want to read their papers to get their exact claims.

Wouldn't you expect that for a larger number of trials, that the true percentage would approach 50%? For millions of trials, one would expect the percentage to be essentially 50%, not a little more than 50%. [/B]

OOK!

Tough question, I am not sure that humans would evenly distribute thier answers across the cloices. So given the non random nature of human choice I would not expect a random answer at all.

Statisticaly, You cant' say that for millions of trial the answer will be 50%, it will approach the limit of fifty percent, but random chance being what it is, you would need a set of runs from a single individual that would be off the random by one standard deviation.

I don't think that any of the researchers got any where near a thousand runs, at least at Duke.

 

[Dancing David]

So what exactly is clairvoyance anyway. They say up to 25% of the population is more sensitive to their surroundings. Stimuli affects them more, and they notice more about everything around them.

I can tell it's going to rain just by how I feel sometimes, but then I go outside and understand why I feel that way. It's perception more than psychic abilities.

Heck, even some people taste more in their food than others. What is super sweet to one person is just barely sweet to another.

'psychic' abilities aren't anything but how a person is built to perceive the world around them. People who say they can talk to the dead are scam artists. The sixth sense is bunk. So I can feel a change in whatever pressure when it's going to rain...but even a barometer can do that.

Thank You EoE!

I feel exactly the same way, most ability to read minds comes from listening to the subtle hitches in a person's speech, by watching breathing martial artists can tell alot.

 

[Eos of the Eons]

You're welcome. I just had to put my two cents in even though most sane people know much of that already.

I liked the comment about the odd number of questions making 50% impossible. Too true.

Anyways-stats are stats and wacky nuts trying to prove wacky nutball stuff with stats is like trying to define baby talk with codes.

This whole thread is interesting in one perspective, but redundant in another

 

[T'ai Chi]

Re: Re: Re: twisted statistics?

I don't think that any of the researchers got any where near a thousand runs, at least at Duke.

Hi David,

I'm fairly sure it is in the millions,... just not entirely sure.

I'll see if I can look it up somewhere.

TC

 

[athon]

You're all missing the point. Statistically, something might be happening, or might not. Given infinite trials, any combination of 'hits' and 'misses' is possible. So in a milion trials, you could technically get 99% being 'hit's'. It's not likely, but it is technically a possible event.

So statistics cannot 'prove' anything. They can support, or point in a direction to explore, but themselves they are little more than red flags which say 'look closer'.

Athon