I just found an interesting article - its very technical but here are some of the points:

(excerpts from article at SCEC site)
We test the hypothesis that anomalous,non-local space-time patterns and correlations associated with recorded events can be detected in real seismicity data years prior to the main shock.The seismicity data employed in our analysis is taken from existing observations in southern California between the years 1932 and the present.Using only a subset of this data covering the period from January 1,1980 through December 31,1991,we compute the probability for finding an anomalous spatial correlation at all sites in southern California over several intervals preceding December 31,1991.We then superimpose on this map the locations of main shocks larger than 5.0 that occurred between January 1,1992 and November 1,1999, that is,the ~8 years following the time interval from which we computed the probabilities.We observe a striking correspondence between regions of increased probability and the location of the recent main shocks,tending to support the results first observed in our simulations.In particular,we note that the epicenter of the recent October 16,1999,M ~7.0 Hector Mine earthquake in southern California occurred at a location that is identified as one of the high probability locations.From the size of the candidate source regions,the magnitude of the possible events can be estimated as well.A likelihood ratio test of the method on both the real southern California seismicity catalog,and a second catalog in which times of events had been reassigned randomly,indicates that the method does find coherent correlated structures in the data.

4.CONCLUSIONS
In summary,we conclude that we have observed systematic variations in seismicity prior to recent southern California earthquakes.Our method employs data from existing seismic monitoring networks as well as a theoretical understanding obtained from numerical computer simulations to identify coherent space-time structures in seismicity.These space-time patterns in the seismic activity directly reflect the existence of correlated structure in the underlying stress and strain fields,a necessary precondition for the occurrence of large earthquakes.Depending
on the nature of future seismic activity in the region,as well as future modifications and extensions of the theory and technique,this procedure may prove useful in analysis of future trends in seismic activity.

*************************************************
The entire article can be read at the link below:

• Earthquake prediction article