Rejecting the Kuriles immediately invalidates the analysis.

Stats 101 says you can't toss out the data that doesn't fit the pattern, then see if the pattern has statistical significance.

What is the significance of the data leaving in the Kuriles?

The irony of all this is that we all EXPECT some small level of correlation of tides with earthquakes (far too small to be useful for prediction, already shown in several meticulous studies of smaller earthquakes, and with the twice-daily rather than twice-weekly tides), but Jim can't get to point of showing there is even a tiny significance, nor can he read and understand the work of the people who have already found it.