From verme003@umn.edu Wed Jun 1 18:51:56 2005
Date: Wed, 01 Jun 2005 15:21:52 -0500
From: Hans Wendt <verme003@umn.edu>
To: rdnelson <rdnelson@Princeton.EDU>
Subject: What role of the experimenter?
Observers have been asking whether the Nelson et al. random generator
results are influenced by the Experimenter, or that individual's
"mind" even serves as an essential conduit, as I understand the issue.
The entwinement may or may not include
currently obscure ("anomalous"?) pathways and dispositions. In the
context it may be productive--even though not ultimate
proof--to look for differences in Experimenter personality and
motivation, changes over time, and other determinants from the
psychological research catalog.
Below I am comparing some of the results obtained among the persons or
"hypothesis sources" in the GCP trials (cf. Table 2)
who made predictions about the connections between some collective
"coherent" state and changes in the entropy of random
event generators. The number of the individuals working in that capacity
is highly unbalanced; about 100 of the 200 or so trials
are due the chief investigator, R. Nelson, the remainder are distributed
among less than 60. Nevertheless, some classifications
are possible, and the parallel statistics may serve as one approach to
the problem at hand. Ignoring for the moment other
confounds a split into two segments also permits a tentative look at
changes over time.
All Z values in Table 2 as of May 3, 2005, have been used. Those
include results from additional processing, secondary
statistics such as variances, and items later removed from the main
analysis because of redundancy or other problems.
There are nine such instances.
A. Events No. 1 - 100 (August 7, 1998 - January 28, 2002).
Nelson only
Others Nelson with others
Others with Nelson
Mean Z 0.4703 (N=47) 0.3343
(38) 1.6803 (3) 0.4837
(12)
S.D. 1.0623
0.9720
0.9057 0.9589
Var. 1.1285
0.9448
0.8204 0.9196
Min. -1.776
-2.085
0.934 -1.123
Max. 3.209
2.428
2.688 2.207
B. Events No. 101- 199 (February 9, 2002 - April 23, 2005).
Mean Z 0.2501 (56) 0.1299
(26) -0.9270 (5) 0.1478
(12)
S.D. 0.9559
0.9425
1.4280 1.2595
Var. 0.9137
0.8884
2.0392 1.5863
Min. -1.723
-2.060
-1.971 -1.599
Max. 2.266
1.710
1.582 1.783
Significance of Differences Set A vs. B, via t-Test, approximate p
(two-tailed):
0.29
0.38
0.01 0.50
While only one difference is significant and between the two smallest
groups at that, all four are in the same direction.
If independence can be assumed despite the involvement of the same
experimenter in most settings this pattern alone has a
binomial p (two-tailed) of 0.125.
Other procedures as well are somewhat suggestive of a time effect as
defined here. Thus the logarithmic combination
of the four p values leads to an equivalent Chi Square of ~ 15 (df=8),
and p ~ 0.05.
The running correlations (r) between event numbers and Z, cumulative
from No. 1 and from No. 101, respectively, are
as follows:
No. 1 - 20, -0.3161 No. 101 - 120, -0.2180
No. 1 - 40, -0.0817 No. 101 - 140, 0.0181
No. 1 - 60, -0.0591 No. 101 - 160, 0.1086
No. 1 - 80, 0.0157 No. 101 - 180, 0.1036
No. 1-100, 0.0774 No. 101 - 199, 0.1462
Only three of ten trends are in the direction corresponding to the four
differences shown earlier, and none are significant.
The A and B group totals are as follows.
Events No. 1-199 (August 7, 1998 through April 23, 2005).
Mean Z 0.3664 (103) 0.2513
(64) 0.0508 (8) 0.3157
(24)
S.D. 1.0150
0.9579
1.7946 1.1081
Var. 1.0302
0.9177
3.2206 1.2279
Min. -1.776
-2.085
-1.971 -1.599
Max. 3.209
2.428
2.688 2.207
Experimenter effects if any, may be confounded by the type of event --an
interaction involving Meditation may be processed
differently from Violence, etc. For a superficial approach to this
issue I examine the patterns of Meditation entries (the most
frequent category, with 39 events when counted liberally) in relation to
experimenter etc. categories. The outcome is as
follows:
(Meditation Events)
Nelson only
Others Nelson with others
Others with Nelson
Mean Z 0.3240 (9) 0.0463
(20) 0.8723 (3)
0.5200 (7)
S.D. 1.2786
0.9454
2.1420 1.3698
Var. 1.6349
0.8938
4.5881 1.8764
Min. -1.707
-2.085
-1.490 -1.324
Max. 2.242
1.479
2.688 2.207
The outcome does not suggest differences among the individuals or the
pairings available, at least not in respect to the
Meditation events.
In conclusion, there is no convincing evidence of differential outcomes
attributable to experimenter characteristics as normally
understood. On the other hand, additional attention should be paid to
the time course of the phenomena seen in the
GCP project.
Hans W. Wendt