My Retropsychokinesis Random Number Generator Meta-Study

[_Ice_Ages_14_Aces_]

[Arouet]

Your hit rate after 171 trials is: 50.25%

You are suggesting that the quarter of a percent above chance is something to get exited above?

[_Ice_Ages_14_Aces_]

[_Ice_Ages_14_Aces_]

Aruoet,

What would be considered evidence in statistical standards is if the P-Value is less than 0.05 since a result like this can be expected by chance less than 1 in 20 (P<0.05), it has nothing to do with the hit rate.

For instance, suppose I tossed a coin 100 times and got 58 heads (58% hit rate where MCE is 50%) Is this hit rate above MCE evidence to you????

[Arouet]

[Arouet]

I'm not a stats guy so I can't debate the finer points with you. But I'm not going to get excited about a 0.25% effect.

[_Ice_Ages_14_Aces_]

Arouet,

Sorry for the indirect reply...

Well....if you were expecting an exact 51% hit rate in my 173 combined study, it will be so highly statistically significant that it will be much more unlikely than winning the lottery....

Anyway, I will publish my meta-study again on friday and we'll see how it'll turn out...

[_Ice_Ages_14_Aces_]

My Retropsychokinesis Random Number Generator Meta-Study Early Update!

In my old meta-analysis that combines all the results of 173 studies, the overall results were statistically significant at P=0.03, which shows fair evidence for psi-hitting.

From Friday-Today, I conducted 112 studies (Total Studies: 285) and here is the overall Two-Tailed, Statistical-Analysis:

Total Studies: 285

N=291,840

Average Chance Expectation: 145,920

Hits: 146,634 (50.24% hit rate & 714 hits above the average chance expectation or in a statistical sense, expected value)

stouffer-z=2.64

P=0.0082 (1 in 121)

Evidence of psi-hitting: Very good!

In the experimental group, the P<0.05 was consistent throughout my 112 studies and my highest statistical significant result out of all the studies conducted at this point is P=0.0029 (1 in 344), which is really close to passing James Randi's One Million Dollar Preliminary Test (P=0.001)

In the control group I also conducted more but not exact as 285 studies and the overall results are still based on what would be expected by chance...

Please leave a comment, thanks!

I will still publish more studies this friday by the way......

[Arouet]

I'm still not terribly exited about a 0.24% effect size!

Unfortunately there aren't that many people (myself included) on this site who are really qualified to evaluate your work here or the stats you've done.. On Skeptiko (where winston is a member as well) there are several from both the skeptical and proponent side. I'm sure they will be interested in this. Maybe you can cross post this there (but do it in the main forum, not the science subgroup: the main forum gets the most attention).

[Jayhawker30]

Say, what exactly defines a chance hit rate?

[_Ice_Ages_14_Aces_]

My RetroPsychokinesis Random Number Generator Meta-Study Update # 2!



This is my PK Study # 346 and as you can see, I was trying to mentally make the bar go to the right (like I always do in all of my PK studies)

Result: In this PK study, I tried to mentally influence the bar to go the right and fortunately it did went to the right in a statistically significant way, i.e. in a way that would hardly hardly happen by chance alone.

Here is my overall Two-Tailed, Statistical Analysis for this study:

N=1,024

Mean Chance Expectation: 512

Hits: 570 (55% hit rate)

Z=3.59

P=0.00032 (1 in 3,125)

Evidence of PK: Very Good!

99% Confidence Interval: 51%-59% hit rates

Getting a result like the one observed in this video can be expected by pure chance 1 in 3,125. In other words, if you run this experiment 3,125 times, you would get a single result like this by pure chance....however, I conducted only 346 studies....

A result this statistically significant would be considered a "Pass" in JREF Preliminary Test

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

Overall Meta-Analysis in Experimental Group

Total Studies: 346

N=354,304

Mean Chance Expectation:177,152

Total Hits:177,853 (50.19% hit rate and 701 hits above the mean)

stouffer-z=2.35

P=0.018 (1 in 55)

Evidence of PK: Good!

99% Confidence Interval: 49-50% Hit Rate

Even though the z-score decreased about 0.27, the result is still statistically significant and therefore, evidence for the PK-Hypothesis....

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

Overall Meta-Analysis of Control Group:

Total Studies: 284

N=290,816

MCE: 145,408

Total Hits 145,437

stouffer-z=0.10

P=0.91

Evidence: None.

99% Confidence Interval: 49-50%

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

Summary

It seems that even though I coducted a total of 346 studies, the overall statistical analyses of all these studies still show consistent high z-scores and evidence of PK from 172-346 studies.

Unfortunately, I don't know how to calculate the overall effect-size of this effect since calculating effect-sizes wasn't covered in my statistics textbook..

sigh, we would know how strong this possible PK effect is if we only knew the damn effect-size !!!!!

[_Ice_Ages_14_Aces_]

Jayhawker30,

The hit rate doesn't matter, it's the P-Values in significance tests that matter. P-Values tells you with a less than 5% (P<0.05), 1% chance (P<0.01), etc. chance that the null-hypothesis is true. Even if you tried to count evidence based on only hit rates, it would be very confusing since a hit rate of 58% in 100 coin tosses isn't evidence, yet a 58% hit rate in 1,000,000 coin tosses is extremely strong evidence..

P-Values are much more accurate and conclusive than hit rates.

Arouet,

Thanks for the advice ...

Sure, I'll publish it on Skpetiko forum (but not right now...)

[Arouet]

[_Ice_Ages_14_Aces_]

Arouet,

The P-Value is the probability, assuming that the null-hypothesis is true, that the observation would occur as extreme or more extreme than the one actually observed. P-Values are completely different than the probabilities in coins, dice, etc.

For instance, suppose I ran 20 significance tests. P=0.05 (1 in 20) tells us there is a 5% chance of a significance test being statistically significant by chance alone, so in this case, statisticians expect about 1 of 20 tests being positive by chance.

[Arouet]

Right, but the one in 20 is an average, you could get that result on your 3 trial, or your 15th. To say that you got a 1/3000 shot in 340 trials isn't not very significant.

Do you acknowledge that the results are edging back towards the mean? It's not going to happen quickly unless you have a bunch of particularly bad runs in a row to match the bunch of particularly good runs you had when you moved from 140 - 171 trials. But your trend seems going back in the opposite direction at this time.

Anyhow, I guess you'll continue running your trials and see what happens. But I really suggest presenting your work in a forum where there are people who can give more constructive criticism than I can.