Canie,
Actually there are only 292 windows that Berkland considers - there are apparently
other factors beyond just new or full moon status. This gives 2336 days within windows
or about 23% of the time. Earthquakes in the four study areas tend to cluster within
these 2336 days preferentially over the other 7679 days between Sept. 1973 and
Dec. 2000 as shown by the statistical analysis.
I have thought this might be due to chance, for example in Central California
the large aftershock sequences of Coalinga and Loma Prieta both fall within
windowed times and in So. California both the Joshua Tree and the Lander sequence
fall in windowed times. I thought by removing the aftershocks for a month after
each of these mainshock, the syzygy effect would disapper. When I did this
the significance remained in the Central California and the Seattle data. The
Los Angeles area data, however showed 23% of events in the windowed periods,
approximate what might have been expected by random chance.
The global 7+ earthquakes had no clustering of aftershocks, so the results
for that data set were not affected by aftershocks or localized clustering in the
same way as those in the western U.S. were.