Can you give an example of this? One where an applicant took the preliminary test, achieved results that had a less than 1% chance to have been guessing statistically, but Randi rejected them anyway because they couldn’t meet their target/agreed to number?
No, how could that possibly slant the test?
Ok, let’s say that I can tell you which object of 10 you put in front of me is blue. You randomly decide where to place the blue item, and then put 9 red items around it. I point out the blue item. We run this test again. Every single time I pick the blue item. So it’s clear that I am able to pick a blue item in a group of objects.
If they determined where to place the blue item via dice roll or someone just putting it wherever they felt like or by randomly tossing it, what could it possibly matter? If I can pick a blue item out of a group, it doesn’t matter how the random sequence to put them in order was generated. The only exception to this is if the testee is aware of the method I am using to place the item - but this would only ever result in a false positive, and never a false negative.
Edit: Sorry if this has been covered. I didn’t notice there was a page 2 when I wrote this response.
Ya know what is hopeless, people trying to pigeonhole elegant and subtle phyiscal systems that are overleaved and multifaceted. Trying to apply a number with the crudest systems and tool to everything with the hammer of logic. Seeing the world as a black and white formula and failing to see the interdictions and paradoxes… using occam’s razor with a literalness that shaves away the skin… Never looking for answers that are outsude of a dogma… frankly, it is obvious we are going through the dark ages of science with no sense of the future and innovation, only measured cynicism and Skeptik Jaundiced Eye.
Because any mechanical method is subject to mechanical bias.
AND they can be faked by a magician. Randi is a magician, even if not a very good one. Randi knows how to fake the selection method used in this particular test. He could make the dowser select any number Randi wants. Faking it is simple, and the basis for hundreds of magic tricks.
ANY mechanical method of selecting random numbers slants the test.
When Putoff & Targ tested Uri Geller, and published their findings in a respected scientific journal, they used a similar method of randomization. This was one of the main criticisms against them.
Likewise, it is one reason why most intelligent people reject Randi.
When you design a test to verify that something has a high confidence of being greater than chance, the higher your claimed success rate is, the fewer tests it takes to confirm that test rate.
So if someone claims a 99% accuracy rate, we can find out in very few tests that they are incorrect. However, if someone claims to be only 5% better than chance, then it takes a whole lot more tests before we can say with statistical confidence that they are operating at a better than chance level. The fewer the tests, the higher the variance, and the bigger the role of random chance, and hence, the more tests we need to seperate results that are closer to random chance.
So what you are probably doing here is taking a test that was designed to confirm an 80% success rate, and fail. But because that person may have succeeded at greater than chance FOR AN INFINITE NUMBER OF TRIALS, you think it’s statistically significant. This is a logical error.
Let me demonstrate this. Let’s say I claim to be able to predict the outcome of a 10-sided die roll. Now, random chance FOR AN INFINITE NUMBER OF TRIALS would get 10% of the predictions right. But let’s say that we run two trials. In one trial, I missed, and in one trial I got it right. I have now predicted the results 50% of the time. And 50% is far greater than chance! Therefore I’m a die roll psychic, right? No, of course not - you can score greater than average random chance over any particular short run of trials.
Which is why when you design a test, you have a confidence interval in mind. You can say after X number of trials, there’s only a 1/1000 chance that you’ll randomly exceed the average chance, or after X+Y number of trials, there’s only a 1/100000 million chance you’ll exceed the average chance.
So even if what you’re saying is true, and a person missed their agreed to purpose of 80% but hit 35%, it wouldn’t prove their ability is greater than chance - because the test was designed to rule out the 80% number to a certain confidence interval, and that number of trials is insufficient to rule out the 35% number at the same confidence. If the person claimed originally that they could only get 35% (when 30% was random chance), then the test would’ve been designed differently with a much greater number of trials to reach the same confidence interval with that number.
Haha, yeah, we’re sure in the middle of the dark ages for science, what with all our progress coming to a screeching halt and silly logic dominating everything.
Make some specific claims here. What are scientists failing to realize due to their skepticism that would otherwise advance the state of science?
NO, don’t say “perhaps”, demonstrate how this could possibly logically produce a false negative. Give me a step by step process whereby your claimed mechanical bias could invalidate the results of the test.
Let’s say someone actually has the ability to dowse. There are 10 boxes, and one has a cup full of water in it. How could the method of choosing which box to put it under - whether it be rolling the dice or just putting the cup wherever you felt like - stop them from finding the water?
If they actually have the power to find the water, they’ll find the water. There’s no way that any sort of flaw in the randomness of placing the water would stop them.
If their power of dowsing was real, then they could pick out the object as easily as I could pick out a blue object in a group of red objects. What order the tester decided to put the red objects in would not matter. I see the blue object, and so I pick the blue object.
I didn’t claim that it HAS produced a false negative. Therefore I have no need to justify the claim.
However, there has been one test that Randi performed. The result was in the dowsers’ favour. This is a false positive, probably due to Randi’s stupidity. And Randi covered the positive result by distortion.
But you are making the claim that this could lead to a false negative. This is not true. If it is, please demonstrate how this could lead to a false negative.
I’m not asking you to prove it has, I’m asking you to demonstrate that it is even logically possible by coming up with an example of how it could happen. What is the logical mechanism behind the false negative?
Please relate to us the exact conditions of the test, including the number of trials. From this we can generate a confidence interval that the results were indeed greater than chance but were ignored, if that is truly the case.
I’m going out for the night, so you’re going to have to find someone else’s attempt to show you the logical errors of your ways you which you will then gloss over and/or dodge… I say this because I know after I don’t respond to your next nonsense for a half hour you’ll say “see! He doesn’t have a response! I WIN FOR THE EIGHTH TIME MUHAHHAHAA”
Can I just point out, by the way, that we are getting further away from the OP’s question.
The point was how to properly conduct a dowsing test. And while I stand by my statement that Randi cheats and is not a model for how to conduct a test, further discussion of him does not answer the question.
This is basic logic 101 and has necesarily nothing to do with Randi. You could ask anyone familiar with the scientific method to come up with a double blind test to test dowsing, and they would do this same thing even if they’ve never heard of Randi before.
You raised the point that throwing dice to determine a random order for the boxes in this experiment works against the testee because dice may have a mechanical bias and not be truly random. The concern you have is that somehow, this aspect of the test could incorrectly conclude that the testee does not have the powers that they claim. This is a false negative. That is to say that they do indeed meet the requirements, but the test falsely declared them not to meet the requirements.
A false positive would be if the testee were declared to have passed the test when they did not demonstrate the powers that were being tested.
Your point was nonsensical in either way. Any psuedorandomness on the part of a die would not give the test either a false positive or a false negative, because it is irrelevant. It doesn’t matter how you come to randomly order the boxes, with one exception. If you follow some obvious pattern that the testee can decode (if first you have it behind box 1, then box 2, then box 3, they may deduce you’ll use boxes 4, 5, and 6 next and then pick them to follow the pattern. If that is indeed how you arranged the boxes, this would result in a false positive. But so long as you have any sufficiently random pattern - whether truly random or psuedo-random - so long as the testee is not aware of a sequence then it cannot bias the result.
Of course the most obvious way to even preclude this possibility is to not let the applicant know which boxes contained the water until after the test is concluded, which I think is standard procedure for these sorts of things.
Essentially your point about dice being psuedorandom is not only irrelevant, but demonstrates that you have no idea what you’re talking about when trying to form logical ideas about this subject.
Isn’t it funny how every single person that you ever discuss this issue with who is familiar with the scientific method tells you that you have no idea what you’re talking about? Do you think it’s a grand conspiracy, or do you think that they’re all wrong and only you are clever enough to see the truth?
You have no contributed to this point whatsoever. You have not added one bit of information that would be useful to a person designing an experiment. All you’ve thrown around is unsubstantiated accusations, and you seem to think that repeating them somehow makes them valid. Whenever anyone has rebuked any of your points, you simply gloss over that fact and then call them “Randifans”.
You are not attempting to engage in debate, or consider what anyone is saying. You are off in your own little world where no one could possibly get through to you, you just keep repeating the same things over and over to convince yourself.
To directly address the OP: The key to any such test is to have complete agreement among all parties in advance as to the terms of the test and what various results or combinations of results will mean.
You also have to realize what statistics actually mean. If you have ten buckets, one containing water, all hidden under boxes, and turn a “dowser” loose, he will have a 1 in 10 chance of getting a positive result. Repeat it twice, and his chances are 1 in 100. At no point does the result prove conclusively that dowsing works. But the more correct results, the higher the confidence that it isn’t random chance.
A better-designed test would bury a jug of water in a one-acre field, plow the field so there’s no way to tell by looking where a hole was dug, and then challenge the dowser to determine the location of the jug within three feet. Then your starting chances are so much lower* that even a few tests become pretty significant.
The so-called dowsers, after missing the jug, could say, “Wait! There’s water 150 feet down under the point I picked! It threw off the results” or “It rained Tuesday and the ground is damp” or whatever. This is why everything has to be clearly stated in advance.
Really, such a test wouldn’t be that hard to set up, if you live in farm country. Find a farmer who’s getting ready to plow a field anyway. Out here, quarter-sections (160 acres) are common. Ask permission to bury some water (make life easy on the dowser and bury a 55-gallon drum of water instead of a little jug). Have someone else keep the dowser being tested away while you bury it just deep enough that the plow won’t hit it. Level the dirt over it carefully. Let the farmer do the plowing. By doing the quarter-section, you’ve turned the odds of finding that drum of water into about one in a quarter-million. Repeat three times with three farms and I’ll be impressed.
One acre is 43,560 square feet. A circle 3 feet in radius around the center of a jug is about 28 square feet. This provides about a 1 in 1,555 chance of getting it right by pure guesswork.
[moderating]
Absolutely true. This is the GQ forum, and your anti-Randi rantings do NOT belong in this thread. Let’s all stop it now.
[/moderating]
GQ is not about debates, nor about you being a winner or loser. Feel free to start yet another thread in GD–or the Pit for that matter–to talk about Randi and/or the paranormal, but unless you’re specifically addressing the OP’s question, stop doing it here.
As for the suggestion to get random numbers from a random number table, that’s no suggestion at all. A random number table is not a source of random numbers, since they had to be put into that table somehow. Saying that you’ll use a random number table means that you don’t particularly care what method you use for random numbers, and you’ll just trust whatever unknown method the table-maker used. Personally, I’d much prefer to use something known to be trustworthy, like a roll of a die.
Also, a random number table creates a small, but entirely unnecessary additional risk of cheating–since the numbers are selected well in advance, if the test subject managed to acquire the table you were using, and memorized it, the additional “randomness” would be offset by the fact that the targets would be entirely predictable.
It’s not a particularly significant risk (and one that there are some ways to fix–such as labeling the targets in a different way for picking a positive and during the test itself)–but there is no risk at all if you use a means of random selection that generates numbers on the spot.
A reasonably fair die is random enough. As long as you don’t roll it for hundreds or thousands of times, the bias will most likely just be lost in the normal stochastic noise anyway.
And even if it’s not perfect, in this type of experiment there is nothing a little bias towards some numbers can really do.
I just wanted to know if you had a 10% chance of getting the correct answer in a single trial, how many trials would you have to run before you started to approach 10% within some x. IOW, what should the statistical margin (I think) of error be? If I run 100 trials, and the dowser gets 11 hits, is that significant? Is 12? What should x be?
It was brought up that it is hard to dig and not find water. A counter claim was advanced as well. Is there any information pro or con?