PROCEDURE
An invitation was sent to colleagues with suitable equipment to do continuous recording of REG data beginning Friday evening and continuing throughout the Saturday events including all of the public ceremonies of Princess Diana's funeral, and continuing for several hours thereafter. The expectation for the experiment was that a standard measure of coherence within the data would correlate generally with global public attention to the events, and particularly with the degree of shared emotional engagement. A maximum effect during the major ceremonies was considered likely, but formal assessment required the specification of a precise analytical expectation even though this was a unique situation. As the events transpired, the decision was made to use the official schedule for Princess Diana's funeral as published by CNN Interactive World News (http://www.cnn.com) on September 4, 1997, to define data segments corresponding to the major portions of the public ceremonies. The schedule began on Saturday, September 6 at 9:08 a.m. when the coffin left Kensington Palace in London. The next event with widespread significance and interest for the public was at 10:45, when the royal standard at Buckingham Palace was replaced by the Union Jack at half mast. At 11:00 a.m. the funeral service at Westminster Abbey started. At 12:00 noon the service ended with a one-minute silence. The coffin was then transported to Althorp, arriving at 3:30 p.m. These four segments of the ceremonies were chosen as well-defined parts that were recognized and attended by the public locally and via worldwide live TV coverage. A more detailed assessment addressing specific elements such as the Earl Spencer's eulogy or the minute of silence ending the funeral could not be accurately specified in all the datasets. Although most of the data were recorded completely automatically, one dataset was taken with the experimenter attending the live television coverage and making concurrent notes and index entries, allowing a subsidiary analysis based on the separate events comprising the funeral service.
The ceremonies for Mother Teresa received less comprehensive coverage, and the schedule of events available to define the pre-planned analysis was simpler, consisting of a three-hour procession beginning at 9:00 a.m., followed by a three-hour funeral ceremony from 12:00 noon to 3:00 p.m., as reported in the New York Times International, Sunday, September 14, 1997.
The raw data in FieldREG experiments are recorded as a sequence of trials each consisting of the sum of, for example, 200 binary digits, with the expected mean value of 100, as if one were electronically throwing 200 coins and counting the heads. Other trial sizes ranged from 400 to 228504 bits. The formal analysis is based on the deviation of the empirical mean from expectation within each segment, and the composite deviation across all the specified segments is compared with theoretical expectation. To quantify the hypothesized effect, a Chisquare statistic is constructed from the sum of squared Z-scores calculated for the segment deviations (Snedecor and Cochran, 1980). This statistic is compared with its theoretical distribution to calculate a chance probability for the deviations found in the data. A standardized format and protocol allows the construction of a "control" distribution based on a resampling process where randomly placed segments of the same lengths as in the active data are drawn repeatedly from the surrounding undesignated data. Distributions based on 1000 repetitions of this process yield a well defined estimate of the parameters of the random data sequence in the absence of an anomalous effect. These empirical control data conform well to theoretical values, indicating that the theoretical Chisquare distribution may safely be used for calculation of the probabilities associated with the active results (Nelson, et al, 97).