Isn't using only NEIC data a problem since they do not report many moderate sized
locally recorded earthquakes. ISC is a far better database which reports about twice
as many events as NEIC. ISC claims to be complete at Mb about 4.5; PDE is doing
well if they are complete at 5.0. Of course, there is the problem that ISC does not report
immediately, but couldn't your program input data other than from NEIC. Especially
for smaller events allowing only NEIC data seems to me not to be the optimal choice.
Predictors could always make the evaluation look bad by showing that a forecast
which was evaluated as a miss using only PDE data was actually a hit using the
local catalog (or other other teleseismic network). What about input from the
GSE (International Data Center) data base. This is by far the most complete and
accurate catalog currently available and can be obtained from IRIS.
I realize that your program will only read NEIS data formats, but it is easy enough
to reformat other catalog data so a better evaluation would result.