DISCUSSION

It is worth noting that although the overall result is driven somewhat by the large value from Edinburgh, this is by no means the only contributor. There is no evidence that the value is an outlier, even though the Chisquare is impressively large compared with most of the others. It is actually well within the range expected for this number of single degree-of-freedom Chisquare estimates.

The overall effect is not overwhelmingly persuasive in terms of statistical significance. However, it is most instructive to consider the effect size in the context of other estimates from field and laboratory studies. An analog to effect size that is particularly informative is the "time-normalized yield". This measure allows sensible comparisons of quite disparate experiments by describing the anomalous yield as a function of the time spent trying to produce it (Nelson, 1994). For the present database it is readily calculated: the p-value for the cumulative Chisquare corresponds to a Z-score of 1.984, and the time-normalized yield (Y) is this Z divided by the square root of the number of hours invested in the generation of the effect. For the 14 five-minute datasets, the total time during which the REG systems were "available" to the hypothesized anomalous influence is an hour and 10 minutes, resulting in a yield, Y, of 1.836. This is approximately nine times as large as the time-normalized yield found in PEAR laboratory REG experiments and is a little larger than the strongest FieldREG effect sizes. Even without the Edinburgh result, the time-normalized yield (Y = 0.643) remains larger than the laboratory effects by a factor of three.