GCP Site Correlations

Dick Bierman made precise predictions for the egg activity around the New Year (Dec. 31, 1998) transition, addressing comparisons between eggs in
Europe and the USA. His assessments also included examination of the intercorrelation of the individual data streams.

Here are the very preliminary results of the analysis comparing US based RNGs (REGs) with European based RNGs around Dec. 31 23:00 UTC (midnight in Netherlands, Germany and Switzerland).

The bad news: The predicted difference in variance during the first minute after midnight (with 1 sec time resolution) couldn't be confirmed  (see graph below for the plot of cumulative chi2).

However: There appears to be a difference developing during the last 5 minutes of the European year due to first an increase in variance in he US-RNG's followed about 60 seconds later by an increase in variance in the Europeans. Furthermore there is a sudden drop in variance of the European RNG's in the second minute after midnight which is recovered in mins 3, 4 and 5. (see graph below which covers the 10 minutes around midnight)

When comparing the compound behaviour of European vs US (Eastern) RNG's at the eastern new year the global picture is similar to the one I have sent for the same comparsion at the European new year (see graph below).

So this runs again against my hypothesis (i.e. there is no notable extra variance in the US - RNG's during the first minute of their new year). However during the 5 minutes preceding this moment the European RNG's are increasing strongly like the US RNG's were doing before europes new year. Then at midnight the cumulative chi-squares come together agian and stay close till minute +03:30 at which point the europeans take off again. I also looked at the correlation between the compound scores within each minute and found no systematic behaviour. Most correlations are negative and the over-all correlation (over 10 minutes) is -0.065 (p = 0.11 two-tailed).

The good news: There appears to be some strange correlation between the European as a whole and the US-RNG's as a whole. 5 and 4 minutes before midnight the correlation is positive, dropping to nothing just before midnight, and turning into negative after midnight with the following correlation coefficients:

-5 mins: +0.138 
-4 mins: +0.213
-3 mins: -0.024
-2 mins: -0.08
-1 mins: -0.09
+1 mins: -0.186
+2 mins: -0.130
+3 mins: -0.164
+4 mins: -0.008
+5 mins: +0.128

I did not download more data than this but it seems to me that this may be a promising avenue. (the 3 minutes after midnight as a whole are significant on itself).

Lots' of work to do.


Correlations on New Year's day

From bierman@psy.uva.nl Sun Jan 3 01:03:01 1999
Subject: PDL-JSASOC: GCP correlations.

Happy new year to all,

Stimulated by James' persistant request to look for correlations between RNG's behaviour I have analyzed the first 6 hours after the new year hit the eastcoast. I made a correlation matrix of the Z^2 per 15 minutes of 8 RNG's some of them in Europe some in the US. There are 28 R's in the matrix so one would expect 1 or 2 significant values just by chance. And of course I could have choosen all kinds of hours to analyze but these are not completely post hoc (I never did this particular type of analysis before)

In the table below I give the significant pairs (N=24):

PAIR R Z-value
1-1000 +0.403 2.14
1-1005 +0.710 4.44
33-1003 +0.730 4.64

No significant negative correlations.


GCP Home