Mike;

> While writing the above, it occurred to me that such an output as you've produced is, in a way, fractal. That occurred to me because my first, rapid reading (combined with Todd's response still floating around in one of my lobes) left me thinking we were talking microquakes, and I also was thinking tiny lat/long squares. But my mistake might be instructive. I had composed half of my response above before I re-read and found we were talking M5 and above, and very large lat/long squares. But the principles involved should be independent of scale. If you think of it in terms of, say, California, small lat/long squares, and microquakes, then, to me, at least, the strangeness of your result disappears.

Quite so. The limits are in the databases. The one I used contains all ANSS from 1973-2005, mag 4+ but if we had one with all Ca microquakes we could do small squares with them. In fact, I think Lowell Whiteside did something similar in his thesis.

> Also - again, I may be missing something, but - wouldn't you expect that around half of all events in a sequence to be overdue in the sense that, if you compute the average time between events, then, by definition, about half of the intervals will exceed that length? Your program immediately flags sequences the moment that interval is exceeded (by a microsecond?). The fact that over half of them are flagged might be because your program actually involves, necessarily, more than a microsecond lapse . . .

Of course. It's like the fact that half the population is below average intelligence. What might help is a frequency distribution for each square so you could tell how likely it is that the quiet will continue. That's a lot more work but it can be done.

What caught my eye was that some of the squares are up to 10 times the average delays.

Roger