Roger:

This discussion triggers the idea (to me) that with regards to the Alan Jones method, "where" the window starts could effect the derived "probability". IOWs, if you shift the window start position (in the Alan JOnes method) you could change the answer.

This suggests that the correct way to get observed probability is to run the calculation mutiple times (by shifting the start position each time) and averaging the results.

Another point is that partial windows should be discarded for evaluation purposes (or maybe keep if it's greater than half a window in duration/size ?).

* * Improvement in the Alan Jones' method for Observed Probability calculation * *

For example, let's say you have a 28 year record of observed quakes and a 10 day window. You would:

1) first truncate the record to be a multiple of the window length in duration (Or use it if it is greater than half a window rule?).

2) Then you do the Alan JOnes method evaluation with the window starting at day 1.

3) Repeat this calculation but first shift the window by another day to the right (and wrap the last window to the start), etc.

4) Repeat this process until the window has been shifted across it's total (width) duration.

5) Now take all these results and average them.

IMO, this would give a better or improved estimate of what the observed probability really is.

Just my .02 worth.