My Retropsychokinesis Random Number Generator Meta-Study

[_Ice_Ages_14_Aces_]

I am posting this topic, so I can inform people about my accumulated data and also for your comments/criticism my meta-study.

Anyway, in this meta-study, I conducted (and still on-going) a Retropsychokinesis meta-study from the website, "http://www.fourmilab.ch/rpkp/experiments/" to see whether or not my mind can mentally influence the outcomes from a Random Number Generator. In addition, I used my own two-tailed, statistical analysis to analyze the overall results of these studies (a.k.a. meta-analysis) simply because I didn't trust the statistical analyses in the website.

Meta-Study since 10/11/2011 (Today):

Total Studies: 141

N=144,384

Hits: 72,423 (50% hit rate, exactly what would be expected by chance)

stouffer-z=1.21

P=0.22 (1 in 4)

Evidence of PK in Random Number Generators: None

--------------------------------

Addressing the file-drawer problem

Unfortunately, one of the most prevalent flaws meta-analyses suffer is the file-drawer problem or publication bias in other words. The file-drawer problem profoundly inflates the overall estimate of the effect in the basis of a meta-analyses, which can easily convince researchers the effect is there when it is in fact a false-positive.

The possibility of the file-drawer problem in this meta-analysis has been eliminated since my meta-study automatically records and publishes all the studies I conducted into my Experimental Log.

Heterogeniety

Another problem meta-analyses suffer from is heterogeniety, which is a meta-analysis that combines completely different studies with different criteria as if they are one large study.

The possiblity of heterogeniety in this meta-analysis has also been eliminated since all my studies are always done on the same standard criteria (i.e. same type of experiments, same # of trials, same time and same everything)

Sampling Bias

It is important to make sure your sampling method is truly randomized; otherwise, it will easily bias your study. For instance, suppose you have a deck of ESP cards and the cards weren't really shuffled very well. Now, let's say you test a test-subject with those ESP cards. With enough feedback, the test-subject might be able to notice that there is a pattern of these ESP cards, which can profoundly inflate the hit rate 25% in the long run....

The possibility of sampling bias in this meta-study is highly unlikely since the P-Value of this study failed to reach the P<0.05 level, so there is no evidence that the sampling method in this meta-study is biased or unrandom in other words...

-----------
Conclusion

So far out of 141 studies, I have found no evidence of psychokinesis in random number generators, but I will still conduct further PK in Random Number Generator studies and I will publish the meta-analyses here every Friday. Hey, maybe if I keep on practicing I might get somewhere interesting .....

I would definately appreciate to hear your comments or criticism regarding my PK Research in Random Number Generators and I would be glad to address it by the way.....

Best wishes

-Ice

[Arouet]

Hey Ice, check you pms!

[_Ice_Ages_14_Aces_]

My Retropsychokinesis Random Number Generator Meta-Study Update!

As you have seen on my previous post, it is quite clear that all the combined results of my 141 studies didn't became statistically significant; however, I conducted a total of 173 studies (N=177,152, hits:89,011)and my overall two-tailed, statistical analysis in this meta-analysis is stouffer-z=2.07, p=0.03: Evidence of PK in RNG: Moderate.

I will still conduct more PK Studies and maybe just maybe this will make the P-Value much smaller, which is more strong evidence of psychokinesis.......

[Arouet]

Ice, you're calling this a meta-study, but what is the data that you're combining? Is it your own personal experiment? Are you doing these trials, or are you merging other people's trials. If so: how homogeneous are these trials?

Maybe you can clarify what you are actually doing here.

[_Ice_Ages_14_Aces_]

Arouet,

Thank you for interrogating my meta-study, I really appreciate it .....

Anyway, the data I am combining is my own accumulated data from my own PK RNG studies. I have never combined any PK study that was not conducted by me. Moreover, the trials conducted in each of my studies is N=1,024, so all the single-experiment trials in my meta-study are 100% homogeneous.. There is no study in my accumulated data that differs from the ones with N=1,024.

To make the point of this meta-study more vivid, I am actually testing the micro-telekinetic alternative hypothesis, i.e. to see if it is possible to mentally influence the random outcomes from a random system.

So far, my statistical analysis of all these studies support this psi-hypothesis.......

[Arouet]

Why are you calling it a meta-study? Are you using different methodologies? If not, isn't it really just one study? (not trying to be picky, just trying to figure out what we're talking about). Are you doing new trials and then just incorporating them into your database?

[Arouet]

I'm trying to figure out the tests on that website. Are you just running trial after trial? Did you decide in advance how many trials you were going to do? You seem to have concluded that after 141 trials there was no affect, you added 30 more and suddenly there was an effect. I'm not sure that's sound ideology. You need to pick in advance how many trials you are going to do. Otherwise you risk just stopping when the variance has it in one direction or another.

It seems that you're supposed to focus on the computer graphic. Have you done a control where you just let it go for the same number of trials as when you're coentrating and comparing the results?

Maybe you can set out a little more particularly what your methodology is. There are a few different experiments on that site you linked to: are you doing all of them, or just one of them.

Anyhow, a real scientist could come up with better critiques.

[craig weiler]

[Arouet]

He said at 141 the hit rate was exactly what would be expected by chance.

[_Ice_Ages_14_Aces_]

[_Ice_Ages_14_Aces_]

Craig,

What exactly is retropychokinesis???

I'm having a hard time understanding this conept...

[Arouet]

Can you take us step by step through what you're actually doing? Is it that for each trial you are staring at that line going back and forth on that website (in one of the 5 or so different options) 1000 times?

And I still don't understand why at 141 trials you would have pretty well exactly expectation, but 30 trials later suddenly have significance? Does that not suggest variance? Your entire significance seems to be based on the last 30 trials! That sounds like variance.

As for the retro: from what I understand: these rng results you are looking at were generated prior to you actually looking at them. When you concentrate on them you are testing whether you can retroactively affect what happened the first time around. So you are basically looking at a recording, not an RNG in real time.

[Arouet]

Also, I get what you are saying about the meta-study, I'm no expert, but I don't think that's what you're doing: it's just one continuous study. I think calling it a meta-study is confusing as it implies that you are combining a bunch of different studies from different people. You're just continuously adding to your study.

Again: I don't think you should be just doing trials and recalculating. You should be doing a set number of trials, decided in advance, Then check your results.

[craig weiler]

Arouet,
In retropsychokinesis, the test subject is attempting to change a result that has already happened, but has not been observed. (No one has examined the results.)

It is not necessary to set the number of trials in advance. It is only necessary to have enough trials to make the results statistically significant. Let me put it this way. If you flip a coin ten times, you can get a wide variety of results, all due to chance. Maybe 8 heads and 2 tails. If you flip a coin 1,000 times however, you cannot get a wide variety of results due to chance. It will be very close to 500 heads and 500 tails. Chance results even out over large samples.

[Arouet]

Have any of the retro experiments done the following control? You have your normal run. Then you have the control run where before the subject runs his trials, someone else reviews the recording first. The subject and the researcher should be blind to which trials had someone observe first or not.