[R-sig-Geo] Question concerning the Kulldorff's method

[Virgilio Gómez-Rubio]

Dear Everton,


> Regarding the rotine that identifies the possible cluster locations, I do
> not know which value should I consider for the fractpop. The spatial
> clustering is very associated to this value and, depending of the considered
> value for the fractpop, I get a different spatial configuration. Is there
> any thumbs rule to define the fractpop?

Not really. I try to keep it within a range of 15-25%. If you have a
very high value of fractpop it is likely that you will test for clusters
with the same configuration of areas and, hence, you will be testing for
the same cluster several times, which is not very efficient.

I have also observed that if the value of fractpop is very high then it
will sometimes find very large clusters (which are difficult to
explain). For smaller values of fractpop, I am able to find smaller
clusters that are nested within that big cluster. Is that what happens
to you too?

Hope this helps.

Virgilio