Dean Radin's Quantum Noise Generator (QNG)

Towards the end Radin talks about a new device consisting of 32 channels of "quantum noise", in which, rather than the approach used in previous micro-PK work and the Global Consciousness Project, the raw noise is recorded rather than being transformed into a stream of bits. This allows the calculation of autocorrelations for each channel (correlations based on different points in time) and mutual information (correlations between the different channels). The records apparently showed anomalous behaviour at around the time when the result of the 2016 US presidential election was announced.

There's more about this in another interview with Radin on a blog called alpha411 (at about 36 minutes):
http://alpha411.blogspot.com/2018/04/par...erica.html

Radin explains that the motivation behind recording the raw noise (every millisecond), rather than a bitstream derived from it, is to avoid applying XOR processing. XOR processing eliminates any bias in the raw signal, and therefore throws away any effect on the relative proportions of 0s and 1s in the bitstream, making it difficult to work out what is causing any anomalous behaviour that's observed. Radin says here that the p values obtained from the presidential election were stupendous - equating to odds of 226 million to 1 for autocorrelation and 81 thousand to one for mutual information.

Of course, taken at face value, this would represent some of the strongest evidence for an anomalous effect ever obtained. For example, it would be far more significant than single events in the Global Consciousness Project. However, it's not clear from these interviews to what extent (if any) the statistical tests were pre-planned. I'd have thought the obvious thing would be to do a series of pre-planned trials using the new hardware, along the lines of the GCP formal hypothesis series. (Though if these levels of statistical significance were reproduced, only a handful of trials would be required.)

Instead, Radin gives the impression that rather than following up this remarkable result, he has embarked on a big project involving the analysis of sentiment in tweets to try to predict shootings in the USA. I find that very difficult to understand.

It looks as though Dean Radin is going to be using the Quantum Noise Generator (QNG) in various kinds of experiments, so perhaps it deserves its own discussion thread.

There's a paper in the new number of the Journal of the SPR (volume 82, number 3, pp. 129-147, July 2018) by Helane Wahbeh, Loren Carpenter and Dean Radin, entitled "A Mixed Methods Phenomenological and Exploratory Study of Channelling". It mainly describes qualitative investigation of some channelers who were observed in action in a four-day study at Mount Shasta, California. But the QNG was also used, with a view to seeing whether it behaved any differently when the channellers were channelling, rather than sitting around not channelling.

The analysis was essentially as mentioned above, but is described in more detail in the paper. An average signal was calculated for each of the 32 channels for each minute. Then two measures were obtained. Firstly, the autocorrelation was calculated for each channel, and then averaged over all the channels. Secondly, correlations were calculated for all the possible pairs of different channels. Then these two minute-by-minute measures were normalised to have zero mean and unit variance, and finally they were added together to produce a combined measure. Then apparently the mean of this measure was compared between channelling and control conditions. (At this point the description of the method loses me, because I don't understand how the mean of the combined measure can be non-zero when it is the sum of two individual measures that have been converted into standard normal deviates. I must be misunderstanding something.)

Anyhow, the combined measure is found to be significantly different between the two conditions, with a p value of 0.024. (Actually, there is quite a lot I don't understand about this analysis, because figures are also given for the two individual measures, and they don't look as though they differ significantly between the conditions. That may be connected with the fact that there is said to be 587 minutes' worth of data for the individual measures in the channelling condition, but 658 minutes' worth for the combined measure.) Then the paper goes on to compare the data for the 18 different entities that are supposed to have been channelled, and finds that for one of them the readings were very different (p = 0.0006). 

The interpretation is that the QNG is detecting "momentary distortions in the temporal and spatial fabric of reality" (or in "space-time"). But it's also made clear that no formal hypotheses were formulated a priori, and that the findings are only to be considered exploratory.

Then these two minute-by-minute measures were normalised to have zero mean and unit variance, and finally they were added together to produce a combined measure. Then apparently the mean of this measure was compared between channelling and control conditions. (At this point the description of the method loses me, because I don't understand how the mean of the combined measure can be non-zero when it is the sum of two individual measures that have been converted into standard normal deviates. I must be misunderstanding something.)

Actually, I suppose this just means that the measures were normalised using an overall mean and variance calculated using data for both conditions.