So, if I understand correctly, the data can be summarized like this:
Reality
A D
Guess A 37 3
D 5 5
Applying Fisher’s exact test gives a p-value of 0.00495, which would be considered statistically significant.
Omitting the two old-looking people who anybody might have guessed were dead gives us this:
Reality
A D
Guess A 37 3
D 5 3
The p-value is 0.0497. The most widely used criterion for statistical significance is p < 0.05, which is of course fairly arbitrary. This just barely meets that criterion.
I had slight doubts about 23,33,46. Anyway, it is over.
Let me assure everybody. May answers were not guesses. The pictures were sorted as per the reaction the dowsing rod showed. Those who have seen the photos would agree that many were not clear. Unfortunately, I do not have a printer online. I tried with somebody’s printer. Since the prints were all appearing smudged, I just took three prints only. Hence reconfirmation with hard copies was not possible. I am not coming out with any excuses.
You would agree that 42 of the 50 were correct. This was not guess at all.
I guess there’s a reason I’m just an old database programmer and not a statistician.
Am I correct in reading this as saying that of the people guessed as alive 37 actually were and 3 were not, and of the people guessed as dead 5 were and 5 were not?
If so I believe that is a correct summary. And thanks!
So this is just barely above the level of chance. In your estimation, for a single test and given the problems mentioned, would you think this result would warrant further testing?
None that I’m aware of, and these were all pictures that I took, or were taken in situations that I was present at and all the pictures came back to me.
There are a few different pictures of a couple of those people on the web, but definitely different pictures. And given that the pictures were cropped to eliminate context (and in fact everything but each person’s head and face), and that there were no names associated with any of them, it would make it tough to google.
Not that I’ve ever made any serious to mask my identity on this board or in my email. And my last name is very unusual in this country (USA). From that I suppose you could search and try to match the bits of information here and there to the pictures I used.
But given the level of success I’d say that didn’t happen, which is about what I expected for a Sunday afternoon lark. I sent some pictures, he dowsed them and posted his results, which appear to be very close to chance.
However, I’m game if pramanujan is to try a more rigorous test.
Yes, there are several problems with this test. Ideally, the pictures would be of people of all the same age taken at the same time, perhaps school photos, to eliminate a likely cause of bias. No information about number of alive and dead should be given. And the analysis should be agreed upon before hand. I don’ know enough about statistics to help on that last part.
If it were a reasonable test of something plausible, sure. In this case? I suppose that’s a matter of opinion.
The statistical test gives a 95% confidence interval for the odds ratio. The odds ratio is the alive:dead ratio for those guessed to be alive divided by the alive:dead ratio for those guessed to be dead. With the two old people excluded, the lower bound of the confidence interval is 1.002676 (it’s expected to be near one since the p-value was just below 5%). So the observation is consistent with a very small bias.
My take on it? Luck alone could explain it, but visual cues from the pictures likely biased the guesses toward correctness. Does this result make me suspect that people can determine dead/living status from photos using divining rods? Nope, not a bit.
Clearly the “diviner” is deluded about what he can do.
I am honest. I have not searched the internet for any clues. As already said, I have just dowsed with the rod. Since some of the photos were not clear, this would have given wrong results.
I am willing to repeat the test with more pictures. Let the photos be a little more clear. It can include photos taken years back too. No wrong if you include one of somebody’s childhood and his recent once, without disclosing.
By the by, this dowsing can be done by anybody. Results will be similar. If anybody can do it, does not qualify for MDC?
The rod I use is a brass one mounted on two bearings each held within the brass handle. The bearings make the rods very sensitive. This avoids any chance of manipulation by hand.
This was designed and got fabricated as per my instructions.
As has already been stated, dowsing claims qualify. You claim that everybody can do it, while they have many years of records where nobody has yet to succeed in a properly administered test.
What exactly does “statistical significance” mean? Is it like saying “outside one sigma of max on a bell curve”? (I am no statistician. As if I needed to explicitly say so. But I know that outside three sigmas is eye-popping.) What kind of result would pure chance give? What would your thinking be if the challenger scores “barely statistically significant” in three or more tests?
As a suggestion for further testing, how about high school yearbook photos from the 60’s, 70’s or 80’s? Pick one year (so the visual difference clues are minimized), make sure the pics are not on the internet*, and have roughly half of each (alive and dead).
*Although I have no idea how an internet image search could reveal the identity of someone in an unlabeled photo.
I’m not sure of the construction you are using, but I’ve never met a rod yet that couldn’t be deliberately manipulated to a desired result if you know a little about physics.
L-rods: the slightest tilt, imperceptible to observers and sometimes the holder, can make them move.
Y-stick: slight tension in a certain fashion, also nearly imperceptible, can make the stick spring violently on one direction or the other. Bearings, by reducing friction, actually make this more likely than using a rigid wooden stick.
No amount of bearings remove this aspect of physics from the situation.
Pramanujan, you failed in the image test. Do you have any thoughts as to why that happened? Although several possibilities come to mind, one of them is that you have no powers as you claimed. Is that one explanation you will consider?
PEAR thought so, but 30 years failed to substantuate their claim.
The problem with accepting such a small margin (51% where 50% would be equal to chance) is that an extraordinary number of tests are needed to provide valid data and extraordinary efforts taken to avoid biases. In the case of the PEAR experiments, where they used an electronic 1/0 random number generator and attempted to influence it with the mind, if the numbers deviated from 50%, it suggested either that the minds were affecting it or the generator was flawed. My money is on the latter.
The point is that they claim to have that level of ability. We’re testing for ability, not statistical significance.
All you can say about a 98%-accurate coin-toss-predictor who scores 70% is that he failed to demonstrate his ability.
You seem to think that if by scoring 70%, he’s shown some ability. I repeat, if he scores 30%, does that show some ability?
Since 70% and 30% are equally likely results statistically, you presumably see them as the same. In terms of demonstrating a 98% claim, they are both failures.
Suppose you test a 98%-accurate coin-toss-predictor and he scores 29%. Would you say that was significant?
If he scored 71%, would you say that was significant?
What range of percentages would not be significant?