I'm sorry Roger but I don't think I'm qualified to answer your question - I just don't want you to think you are being ignored.
I did find this reference:
The null hypothesis should include well know spatial variations and temporal clustering of seismicity (Kagan & Jackson, GRL, 1996).
The formulation of the null hypothesis for large earthquakes (M>7) is relatively easy, since the clustering of these earthquakes is weak, thus the Poisson process can serve as a null hypothesis. However, even in this case, a spatial inhomogeneity of earthquake epicenters presents a serious challenge to the verification process (Kagan & Jackson, JGR, 1995). For moderate earthquakes (7>M>5), the simple Poisson process is no longer an acceptable null hypothesis. These earthquakes are strongly clustered, the best known example of clustering being aftershock sequences. Thus, the null hypothesis should include clustering as its major feature.
(found at: http://www.seismo.demon.co.uk/Nov7th/kagan.html )

Canie