Originally posted by jzs
I'm asking if the RNG output is statistically different than what one would expect by chance.
Different from what one would expect by chance, assuming what null hypothesis?
Specifying a null hypothesis is very important. It is what allows us to calculate what we would expect to see by chance, and it is what gets disproved (more or less) should we not see what we expect.
If we take as our null hypothesis that our RNG produces perfectly independent bits, each having pefectly equal probability of being 0 or 1, then it's easy to calculate what we expect to see. But then, if we don't see what we expect, all we've shown is that our RNG isn't perfect. Big deal. It's already well known that real RNGs aren't perfect, and in any case we've shown nothing about the existence of global consciousness, whatever that is.
If we want to have any chance of demonstrating global consciousness, our null hypothesis needs to be that global consciousness doesn't exist. Now how are we supposed to calculate what we expect to see from our RNG, based solely on that null hypothesis? All sorts of RNG imperfections are possible even in the absence of global consciousness, and therefore seeing any of those imperfections in our RNG tells us nothing about whether global consciousness exists.
Show me their exact claim from that page that supports your claim that they say "RNG's are not necessarily truly random over a long period anyway".
That page recommends "to XOR the random byte with a pseudo random byte. In that case the resulting bytes will even behave properly for higher order bias effects." If the RNG were necessarily truly random, there would be no need for this recommendation.
That's why you look at the data they produce, Claus Lite, to see if they are out of wack or not.
How do we distinguish out-of-whackness that is due to global consciousness from out-of-whackness that is due to more mundane causes?
What the GCP people do for a 'control' is compare the data on a day that is in the formal hypothesis registry to a neighboring day that is not in the formal hypothesis registry.
Where do they say this?
How do they get a specific p-value from such a comparison?
The impression I got from what I've read on their site is that they may look at a few other days just as a sort of informal sanity check, but when it comes down to actually calculating a p-value, they calculate it based on the assumption of an ideal RNG.
Seeing nearly-ideal behavior on a few days does not guarantee that the RNG's behavior is consistently ideal and therefore that any departure from idealness on other days is necessarily due to the effects of global consciousness.
How about examining the data? That seems reasonable.
Yes, of course.
What should we look for in the data?
What are we justified in concluding, if we find it?
The
Global Consciousness Project is called "The Global Consciousness Project," not "The Random Number Generators That Behave Somewhat Nonideally But We Don't Know Why Project."
