Chris;

I agree, the results do look strange.

But if you think about it, the small intervals are a small population...

No, that won't work. If we assume a reasonably uniform time distribution the recent numbers should be the same (or larger) then the more distant numbers.

Am I sure the program is correct? No, I'm never sure of that, particularly when it's dealing with so much data I can't watch what it's doing.

Here's the logic;

I have a list of 47k+ quakes in chronological order.

I read in 30 days worth of quakes starting with the first one. I check the distances from the first one to all the rest. If the distance is beteween 500 and 5000 km I compute the time interval between the first quake and whatever one I'm looking at. That difference is the array index (0 to 29) of the cumulative graph and that cell is incremented by one.

Then I move the quake array up by one (eliminating the first one) and read in some more until the last one is out of date range. The new top quake becomes the main quake and the process starts over.

This continues until the file is exhausted.

Looks good on paper anyway.

Roger