LISA: Flag for islands?

[GoogleCodeExporter]

If this takes much time to do, let's skip it. If it's easy, let's add a flag 
that appears if the weights file contains islands to warn users not to create 
local Moran maps for shape files with islands since the significance levels for 
the islands will not be computed correctly. Users can either exclude the 
islands from the shape file in a mapping program outside of GeoDa or use 
distance weights without islands.

Original issue reported on code.google.com by jkoschin...@gmail.com on 7 Dec 2010 at 9:39

[GoogleCodeExporter]

This won't be too difficult to do, especially with the use of the Boost Graph 
Library.  However, I need to know what qualifies as an "island" in the case of 
non-symmetric weights.  Do we say that an undirected graph (ie non-symmetric 
weights) has no islands iff it is weakly connected, or do we require strongly 
connected?  My hunch is strongly connected, but I'm not sure.

Original comment by mmcc...@gmail.com on 12 Jan 2011 at 8:23

• Changed state: Accepted

[GoogleCodeExporter]

This has been clarified.  An "island" is really an "isolate."  An isolated 
observation is any observation that has no neighbors, although it may itself be 
a neighbor of another observation.  We should add another category to the LISA 
map to indicate isolated observations which do not have valid LISA values.  I 
imagine that isolates need to be excluded from Moran's I calculations as well.

Original comment by mmcc...@gmail.com on 14 Jan 2011 at 7:09

[GoogleCodeExporter]

Resolved by SVN revision 1233. This will appear in 0.9.8.19 release.  A new SVN 
issue has been created to add a new class of LISA clusters for observations 
with no neighbors.

Original comment by mmcc...@gmail.com on 1 Feb 2011 at 10:36

• Changed state: Fixed