Discuss 73. Skeptoid’s Brian Dunning Finds Global Consciousness Project Lacking

[Miguel]

Originally Posted by alextsakiris

Hi HW... given the many hours of interviews and pages of transcripts this show has given to the GCP, and the inability of our skeptics to mount any serious challenge to the findings, I think it's ok to ask skeptics to bring something to this debate other than rhetoric.

Crunch some numbers and get back to us.

There is no way of conclusively proving cherry picking by crunching numbers.
When you see a statue you cannot reconstruct what the boulder looked like it was chiseled from. One can only assume that a statue must be man-made based on the fact that boulders of that shape don't occur naturally.
The observation that results 0.0 to 0.4 opposite the predicted direction are missing from their registry supports the notion but it can never be conclusively proven.

There also is no way to conclusively prove multiple analysis by crunching numbers. That is did someone maybe run many different analyses and keep the ones they liked One can only investigate what happens when the parameters are slightly changed. If the result depends greatly on chosing just the right "recipe" then this may support the notion. However, it does not prove anything.
Just that was done for Radin's 9/11 analysis. It was indeed shown that his oh so significant results relied on some very precisely chosen parameters.
He afterwards admitted to running multiple analyses and publishing the best.
If done by a reputable scientist this might regarded as a serious case of misconduct.

So basically:
-It is impossible to prove there is nothing in the data. (Can't prove a negative)
-It is impossible to prove that the results are entirely due to statistical shenanigans. It might at best be possible to elicit further admissions of such.


What could be done is another analysis, done properly, to see if a good analysis gets similar results.
Anyone up for that?

[alextsakiris]

Originally Posted by Miguel

There is no way of conclusively proving cherry picking by crunching numbers...

There also is no way to conclusively prove multiple analysis by crunching numbers.

So basically:
-It is impossible to prove there is nothing in the data. (Can't prove a negative)
-It is impossible to prove that the results are entirely due to statistical shenanigans. It might at best be possible to elicit further admissions of such.

You're just not correct.

You have a large database of supposedly random data. You can randomly select portions of the data (hour, day, whatever) and test to see if it conforms to your predicted randomness (i.e. do the coin flips come out 50/50). This will tell you if the system is "random". This is your control data.

You can then pick other portions of the database that are predicted to show global consciousness coherence. Compare the two groups and see if there is an effect.

Just that was done for Radin's 9/11 analysis. It was indeed shown that his oh so significant results relied on some very precisely chosen parameters.
He afterwards admitted to running multiple analyses and publishing the best.
If done by a reputable scientist this might regarded as a serious case of misconduct.

this is a gross mischaracterization of what happened... actually it's just plain wrong. Back it up, or retract.

[Danny_D]

Originally Posted by Miguel

What could be done is another analysis, done properly, to see if a good analysis gets similar results.
Anyone up for that?

Isn't the raw data available on their web site? Why dont you have a go at it and do a proper analysis.

[FastEddieB]

...and test to see if it conforms to your predicted randomness (i.e. do the coin flips come out 50/50).

If something can be predicted, doesn't that make it non-random by definition?

Like I've said before, 50/50 is the most probable outcome, but getting exactly 50/50 is improbable.

[alextsakiris]

Originally Posted by FastEddieB

If something can be predicted, doesn't that make it non-random by definition?

Like I've said before, 50/50 is the most probable outcome, but getting exactly 50/50 is improbable.

Obviously, you're not going to get 50/50... remember they sample 200 times a second (i.e. 200 coin flips every second). But, over several minutes/hours you will have a lot of samplings... they will either conform to a bell curve or not. Come on guys this is basic statistics.

[Miguel]

Originally Posted by alextsakiris

You're just not correct.

You have a large database of supposedly random data. You can randomly select portions of the data (hour, day, whatever) and test to see if it conforms to your predicted randomness (i.e. do the coin flips come out 50/50). This will tell you if the system is "random". This is your control data.

You can then pick other portions of the database that are predicted to show global consciousness coherence. Compare the two groups and see if there is an effect.

I don't see how that contradicts what I said.
If such an analysis came back empty, would that show that there is nothing in the data?

this is a gross mischaracterization of what happened... actually it's just plain wrong. Back it up, or retract.

I refreshed my memory. These are the facts:
-The "formal" analysis of 9/11 was indeed done by Radin and Nelson together not by Radin alone as I implied. (Summary of Results, Intro)
-The results have been shown to rely on a "fortuitous" choice of parameters. (http://www.lfr.org/LFR/csl/library/Sep1101.pdf)
-Nelson has admitted to running various analyses before making the "prediction", however he did admit so from the start rather than only after prompting. (Formal Analysis, September 11 2001)

I was mixing this up with an analysis Radin did for the previous New Year. Here, too, it was shown that the analysis depends crucially on chosing certain parameters. And here, Radin admitted to running many analyses and picking the best. (Results, Y2K)

[Miguel]

Originally Posted by Danny_D

Isn't the raw data available on their web site? Why dont you have a go at it and do a proper analysis.

It is a lot of work for no expected pay-off.

If I do that and get the expected results it would simply be dismissed. After all, if the widely published criticism is not convincing then another null result won't be either.

The least I'd expect is that some believer confirms that the protocol is fair and proper and should yield results.

[Paul C. Anagnostopoulos]

Originally Posted by alextsakiris

Obviously, you're not going to get 50/50... remember they sample 200 times a second (i.e. 200 coin flips every second). But, over several minutes/hours you will have a lot of samplings... they will either conform to a bell curve or not. Come on guys this is basic statistics.

They will always conform to a bell curve. The question is whether you pick a period of time that happens to be an outlier on the curve.

~~ Paul

[alextsakiris]

Originally Posted by Miguel

I don't see how that contradicts what I said.
If such an analysis came back empty, would that show that there is nothing in the data?

Do you and Paul really not get this or are you just messing with me?

[Miguel]

Originally Posted by alextsakiris

Do you and Paul really not get this or are you just messing with me?

I am not messing with you. I mean what I wrote.
I can't talk for Paul but I don't see anything in his recent postings in this thread that would make me think that the situation is different with him.