The Global Consciousness Project

Well, thanks again for looking at that. It was never to be expected that we'd get to the bottom of it in this thread, after all.

Regarding further investigations by the GCP people - my impression is that it's just Roger Nelson running things now. The network is still operating (albeit with a smaller number of RNGs than it had at its peak last decade) and there is an active Facebook page, where they are continuing to look at individual events. But the formal pre-registered series has ended, so I think it's being done on a purely post hoc basis now. Roger Nelson is also planning to publish a book on the project.

I think Peter Bancel feels he's nailed it down to an experimenter psi effect, and has moved on to other stiudies. He gave a talk as part of the 2017 PARAMOOC course, which I think contains analyses in addition to those he has published. I presume that's still available to view for free after a registration procedure.
https://carlossalvarado.wordpress.com/20...mooc-2017/

I started to do a bit of analysis a while ago, but there really are a lot of data to handle. I got as far as breaking down the contributions into hourly totals for each pair of RNGs, with the idea of trying to see if any subset of the data can be identified as the source of the effect. I haven't had a chance to do anything on that for a while, though. I'll try to make a bit more progress if I can get a chance.

So the suggestion I made right back at the start, that this signal could be something that is also correlated with something to do with human behavior - energy usage - to which these devices are vulnerable is absolutely valid. *If* we accept that the researchers are choosing major events to analyse in a fair, unbiased way, then that looks to be a possible reason to consider.

I think you are on the right track in several ways, but I am unconvinced that the "hidden message" (if there is one) lies in energy usage. 

I mean, look at logistically how that would manifest. It would be about people getting up and turning on the AC or putting on the kettle for a cup of tea. In my mind, that is MUCH to gross of an input mechanism for a system like this. There isn't enough nuance, enough complexity, enough resolution, in that type of design. 

Also as I said some time ago, energy usage translates mostly to demand, not noise. If power companies are doing there job well (admittedly not always the case, nobody's perfect after all) they will vary generating capacity to match demand. If that is managed well, there will be no real power quality change. 

So yes, I believe there is quite possibly signal embedded in the noise, but I don't think it is due to demand because that just doesn't fit correctly with the characteristics I'm seeing. 

The good news however, it really doesn't matter the source of the signal, just so we can extract it. 

But to your point (I think), understanding the way the signal is injected into the data stream, would help immensely in decoding the message...  

OK so finally on this point, my intuition tells me this:
it may be harder to figure out how the signal is injected, and easier, (via trial and error combined with brute force engineering and data analysis techniques) to tease the signal out of the noise. And maybe when we get a clearer look at the characteristics of the signal, we can back our way into how it gets there in the first place. That's that way many very complex engineering/science problems have been solved in my past experience. The idea being: to grab onto whatever slim thread of coherent information and can find, and gently, through whatever means you can dream up (direct cause and effect don't need to be figured out or relied on at this stage), untangle it to the point where it will sit still long enough for close inspection. Then, from the close inspection, you can glean the data necessary to be smarter about your approach. This is where cause and effect ARE figured out, and can now play a part in shaping/informing a smarter, more appropriate system design.

That's how I look at it purely from an engineering/system design standpoint. Hope it makes sense to others.

One odd feature of the GCP results is that Bancel analysed the correlations between pairs of RNGs according to the type of device. The results, in Table 1 of this paper -
https://www.researchgate.net/publication...xploration - are shown below.

[Image: REGPairs.jpg]
The differences between the three kinds of pairs are not statistically significant, but it's interesting that the correlation is weakest between the pairs of Mindsongs, and strongest between mixed pairs, where one device is Mindsong and the other is Orion.

Thinking about the scope for statistical artefacts, I would have guessed that the scope was probably greatest for the pairs of Mindsongs (because of the complicated XOR mask and the fact that only part of the mask was used for each sample of bits), and that the scope would be weakest for the mixed pairs.

One odd feature of the GCP results is that Bancel analysed the correlations between pairs of RNGs according to the type of device. The results, in Table 1 of this paper -
https://www.researchgate.net/publication...xploration - are shown below.

[Image: REGPairs.jpg]
The differences between the three kinds of pairs are not statistically significant, but it's interesting that the correlation is weakest between the pairs of Mindsongs, and strongest between mixed pairs, where one device is Mindsong and the other is Orion.

Thinking about the scope for statistical artefacts, I would have guessed that the scope was probably greatest for the pairs of Mindsongs (because of the complicated XOR mask and the fact that only part of the mask was used for each sample of bits), and that the scope would be weakest for the mixed pairs.

That makes initiative sense to me.

Did they identify geography as well in these data?

Can you help those of us who are less familiar with the nomenclature of statistics, to relate these figures, and visualize how much of a correlation there is in terms of probability (ie- 1:100, 1:1000 etc) or some other more "accessible" way for a layman?

Did they identify geography as well in these data?

Can you help those of us who are less familiar with the nomenclature of statistics, to relate these figures, and visualize how much of a correlation there is in terms of probability (ie- 1:100, 1:1000 etc) or some other more "accessible" way for a layman?

Yes - the longitude and latitude of each RNG should be available at the GCP website. You can also see a map here (they tend to be most concentrated in North America and Europe):
http://noosphere.princeton.edu/egghosts.html

The correlations are very weak (as indicated by 10^-5 at the top of the column), but as there are billions of data points, they turn out to be statistically significant (apart from the Mindsong-Mindsong). The corresponding p values (according to this online calculator - http://www.socscistatistics.com/pvalues/...ution.aspx - selecting the one-tailed test) are:
Mindsong-Mindsong, p = 0.067 (not significant at p = 0.05, partly because the numbers of Mindsongs are smaller than the numbers of Orions).
Orion-Orion, p = 0.00034.
and for the others the p value is so small it returns only p < 0.00001. I'd have to find a better calculator, but obviously those values are extremely statistically significant.

You could go mad trying to find patterns in the existing GCP data Chris.

We all go a little mad sometimes.

But seriously, if it's a statistical artefact, it shouldn't be a hopeless task trying to track down where it's coming from. If it's psi effect I agree that looking at subsets of the data is unlikely to tell us much (or much more than Bancel's analysis already has told us).

Yes - the longitude and latitude of each RNG should be available at the GCP website. You can also see a map here (they tend to be most concentrated in North America and Europe):
http://noosphere.princeton.edu/egghosts.html

Peter Bancel looked at whether the strength of correlations between pairs of RNGs decreased with the distance between them. He found it did for "minor" events, but didn't for "major" events (though obviously there's an element of subjectivity in this classification).