August 11 Eclipse of the sun, Individual Eggs
The eclipse on the 11th of August, 1999, followed a path across
Europe and the middle east from 09:30 to 12:36 UTC.
Seven GCP host sites were actually in the
path as shown in this map and table:
The following figures examine the behavior of the Eggs
which were actually in the path of the eclipse using the raw trials data. Five of the
seven devices show a consistent
positive correlation of deviations in the eclipse data compared with theoretical
expectation and with control data from the subsequent day.
When all the individual eggs in the path are considered, the total deviation leading up to the
eclipse peak and during the whole period of partial and full eclipse is quite strong.
Some eggs are independently significant, and the concatenation for the time leading up to
the peak shows a difference from expectation with a probability of 3 parts in 10000. For
the whole time of the eclipse, the overall probability is 8 parts in 1000. For the same
time-periods on the following day (control) neither of these calculations
approaches significance. See the following table for details.
If we look indiscriminately at the whole period of the eclipse, rather than the
exact time the egg was in the path, the results are a bit different. We won't
put all the figures on this webpage, but the following email exchange gives a
reasonable summary of what happens to eggs not in the path, and to those in the
path if the analysis does not use the exact timing.
Date: Wed, 29 Sep 1999 00:22:39 -0400 (EDT)
To: George deBeaumont
Subject: Re: revisiting the eclipse
On Tue, 28 Sep 1999, George deBeaumont wrote:
> Here's the cumulative deviation graphs for the individual eggs.
> The graphs are for the entire eclipse period of 09:30 to 12:36 UTC.
Yes, that makes sense. I added up the eclipse eggs and got worried
because the result was different from your earlier table -- but then
I realized that was tailored to the actual time the egg was in the
eclipse path, a shorter time. The result this way is p=0.110,
compared to p=0.008 for the tailored version. If we take this as a
directly meaningful comparison, it suggests the optimum analysis must
take account of the specifics: a general eclipse period is too broad.
However, I think this might be overinterpreting, even though sensible.
> My first impression: the eggs in the path of the eclipse are
> significantly less deviant when looking at the entire period. Gosh, it's
> a subtle phenomenon we're chasing!
Ah yes. My numbers confirm your impression.
> Non eclipse eggs look (for the most part) unimpressive. But I'm not sure
> how to quantify this. Anyway, more for you to ponder... look forward to
> your thoughts!
They do appear to be somewhat less impressive. They accumulate a
positive quantity, but the probablility is 0.227. The literal
interpretation, in the mode described above, would be that they are doubly
disadvantaged. They have no time in the eclipse path at all.
When all eggs are considered, the p is marginally significant at 0.092.
I also did a rough-and-ready approximation of your other analysis,
using the first half of the period in analogy to your "peak" computation.
This is estimated from the graphs, and yields p=0.093 for the eclipse
eggs, p=0.243 for non-eclipse, and p=0.089 for the combined data.
The pattern is the same, showing a somewhat more impressive outcome for the
I did one more calculation, namely the proportion of positive outcomes
for the eclipse eggs (6 of 7, approx Z=1.9) the non-eclipse eggs (9
of 13, approx z=1.4) and all eggs (15 of 20, approx Z=2.2). These
proportion values suggest slightly stronger results, but again with the
Yes, we do have a subtle beast on our hands. Enough indications of
something there to preclude a conclusion that it is all chance
fluctuation, but not enough to go shouting from the rooftops about
global consciousness captured in our net.
(September, 1999, RDN, Figures by George deBeaumont).