Dear R Users;
I have been using the space-time K-function analysis to detect the existence
of clustering in both space and time. I wish to ask and know what the
minimum number of cases should be for such analyses to be valid. I have
data with 23 cases and don't know how careful i should be in the
interpreting the results.
Secondly, i wish to find out if there is any bond between the results from
the SatScan software and Splancs. In an exercise, SatScan suggested the
presence of space-time clustering but Splancs did not present significant
results. This might have been due to the small sample size or the way in
which the spatial distances and times were specified for the analysis. Using
the Burkitts data supplied with the software and changing the spatial
distances and times for the analyses, affected the significance of the
results. Before or after certain spatial distances, the results were no
more significant, also certain specifications of the times ruined the
analyses. A follow-up question to this is if there is a paritcular techniqe
for chosing the spatial distances and times for the analyses.
My approach of space-time clustering is to use the space-time K-function
analysis to check for the presence of clustering. Once present, a search
for their locations and tests for their significance could be carried out
using SatScan. If not, then the plots of the spatial K-function and the
temporal K-function could be used to drive the SatScan analysis into the
spatial or temporal domains. Has anyone used this same approach and found it
correct?
All help will be warmly embraced.
Thanks in advance for your help.
Abatih
KENNIS IS MACHT
