Skepticism and the statistics of GCP

[latent aaaack]

There doesn't seem to be a recent thread about this. The Global Consciousness Project's overall summary of results, August 1998 to March 2005, to be found on this page http://noosphere.princeton.edu/ , shows a probability of 1.869x10^-6.

Can someone illustrate that this an example of statistics being misused to support an invalid claim? This board is accustomed to shooting down verbal arguements and logical errors but it's perhaps at least equally important to show how math can be used to decieve too.

[Cuddles]

Hmm, I didn't manage to read as far as any maths, but I don't think it's actually needed. The second paragraph claims they used "random event generators". If they are random, then there can't be any pattern in there, by definition. If there is some kind of pattern then the events aren't random and their entire analysis is completely pointless.

[latent aaaack]

Sorry the link doesn't work, click on 'primary results' to get to the page I'm talking about.

[CFLarsen]

http://forums.randi.org/showthread.php?t=35352

http://forums.randi.org/showthread.php?t=33538

http://forums.randi.org/showthread.php?t=42900

[latent aaaack]

I'm interested in viewing this paper if someone can link to it or summarize.

http://forums.randi.org/showthread.php?postid=884316#post884316

"Since PEAR are involved in this, I would refer you once again to their own paper admitting loss of effect following more extensive analysis of their own data. This may be a recurring theme with them..."

[latent aaaack]

http://forums.randi.org/showthread.php?t=35352

http://forums.randi.org/showthread.php?t=33538

http://forums.randi.org/showthread.php?t=42900

Correct me if I'm wrong but after reading your article and some commentary I took that the GCP results are viewed as not valid because 1 they 'select their data' and 2 because even the claim that the project is capable of producing both random numbers and anomalousness is disputed.

I don't accept the first criticism but maybe I'm not understanding the criticism.

Data selection here seems to mean that they don't count or explain non-random data when it occurs with no corresponding event and they don't count or explain events that don't produce non-random data.

I'll use the example of the 9/11 data. Despite it representing 1/200 odds it was criticized as being flawed because of similar results being produced in the days before the attack for no reason. One would expect that such non random results would show once in a while only by random chance even if the GCP hypothesis were true. Are you sayng that results representing 1/200 odds are produced pretty much constantly by the project, which means there's in fact no significance when it corresponds with an event? I don't think that's what you're saying but I'd like clarification.

If by data selection it's meant that they aren't counting major events that don't produce non random data I don't think that computes. For example, the criticism would be that an event which corresponded with with a probability of 1/200 in nonrandom data is not valid evidence of GC because maybe they didn't count 199 other major events that had no effect on their data. Having looked at their primary results page it looks like they did count every significant news event so I await further clarification of the points of skeptical criticism on the GCP.

[CFLarsen]

Correct me if I'm wrong but after reading your article and some commentary I took that the GCP results are viewed as not valid because 1 they 'select their data' and 2 because even the claim that the project is capable of producing both random numbers and anomalousness is disputed.

I don't accept the first criticism but maybe I'm not understanding the criticism.

Data selection here seems to mean that they don't count or explain non-random data when it occurs with no corresponding event and they don't count or explain events that don't produce non-random data.

I'll use the example of the 9/11 data. Despite it representing 1/200 odds it was criticized as being flawed because of similar results being produced in the days before the attack for no reason. One would expect that such non random results would show once in a while only by random chance even if the GCP hypothesis were true.

Why?

Are you sayng that results representing 1/200 odds are produced pretty much constantly by the project, which means there's in fact no significance when it corresponds with an event? I don't think that's what you're saying but I'd like clarification.

If by data selection it's meant that they aren't counting major events that don't produce non random data I don't think that computes. For example, the criticism would be that an event which corresponded with with a probability of 1/200 in nonrandom data is not valid evidence of GC because maybe they didn't count 199 other major events that had no effect on their data.

That's exactly it: If they get 200 blips, and one happens (with very liberal interpretation!) on 9/11, how do they know that the 9/11-blip was evidence of a global consciousness?

Having looked at their primary results page it looks like they did count every significant news event so I await further clarification of the points of skeptical criticism on the GCP.

If you are talking about this page (http://noosphere.princeton.edu/results.html), they clearly are not counting every "significant news event": They have done Earth Day 2001 and 2006, but not 2002, 2003, 2004 and 2005.

What the heck is a "significant news event" anyway?

[latent aaaack]

Why?



I meant to say results (like a 1/200 event) that are improbable should occasionally occur just by random chance. That alone can explain why such a result was observered in the days before the attack, that it was just a random spike. The only arguement I can think of against it being attributable to random chance would be if what was called an unlikely 1/200 event was actually not, because such events happen way more often than 1/200, like every week. I don't know that to be the case and I'm asking if that's what skeptics are claiming.


That's exactly it: If they get 200 blips, and one happens (with very liberal interpretation!) on 9/11, how do they know that the 9/11-blip was evidence of a global consciousness?


Their claim is that major events cause improbable results to appear in the data. Assuming they are using every major event for now, when a major event happens, that would constitute an immediate claim by them, that an improbable result should be seen in the data around it. If improbable spikes are seen around the event then that constitutes statistical weight in favor of their claim, the way that you'd measure someone who claimed to guess random numbers.



If you are talking about this page (http://noosphere.princeton.edu/results.html), they clearly are not counting every "significant news event": They have done Earth Day 2001 and 2006, but not 2002, 2003, 2004 and 2005.

What the heck is a "significant news event" anyway?

I agree that they should use more stringent methods to include events but on that page it says that the values for those events were in the predicted direction so were not actually improbable, if I understand what they mean by 'predicted direction'.

[latent aaaack]

Have I correctly described the skeptical position on the subject of the statistics of the GCP? The skeptical takes on this that I've read do not seem to explain why the GCP's claim of a 1.869x10^-6 probability is invalid, only that certain parts of their methodology are suspect.

[T'ai Chi]

The skeptical takes on this that I've read do not seem to explain why the GCP's claim of a 1.869x10^-6 probability is invalid, only that certain parts of their methodology are suspect.

But to a strong believer in the skeptical movement,

"certain parts of their methodology are suspect" = "is invalid"

Just casting any doubt is enough.

(They are ignoring, of course, that metholody being suspect happens in science all the time)

[CFLarsen]

But to a strong believer in the skeptical movement,

"certain parts of their methodology are suspect" = "is invalid"

Just casting any doubt is enough.

(They are ignoring, of course, that metholody being suspect happens in science all the time)

The Chronology of T'ai Chi's "Organized Skeptical Movement" claim (http://forums.randi.org/showthread.php?t=68791)