I agree, Canie, that the probabiity that this quake would occur in the time
frame given was fairly high compared to other predictions Don has made.
That, however, does not invalidate the prediction, it merely makes it more
likely that it could have occurred randomly.
It is important to hold everyone to the same standard, however. If Don
or some other predictor has been relatively successful, future predictions
should not be judged on the basis of past accomplishments. Each prediction
should be judged as an individual entity.
So, here are the statistics using the Jones method preferred by E.P.R.
Since 1933 there have been 1344 earthquakes which match the parameters
of Don's forecast (i.e. they are within 40 km of 35.7N 117.8W with Ml 2.8-3.9).
The length of the forecast is 5 days Aug 23-28. Good data is available
for this region in the stated forecast range from 1933 - 2001. This is
a period containing 4895 5-day periods. Since many events in the region
occur in swarms, the Jones method which attempts to account for
clustering is the preferred technique for determining the probability
of a correct random forecast for the area.
Of the 4895 5-day periods, 594 contain at least one earthquake of
Ml 2.8-3.9. The random probability of "guessing" a 5-day period
correctly is therefore 594/4895 = 0.12. The odds of a correct guess
then are about 1 chance in 8. Better chances than most of Don's
predictions, but still by random guesswork, he would have been wrong
7 out of 8 times.
These odds may be slightly high because a swarm did occur there
about a month and a half ago, but in any case, the prediction
is still considerably better than random.