Is it kosher to analyze dataset containing non-contiguous polygons for clustering with Optimized Hotspot Analysis?

Question

I am running a hotspot analysis on polygons assessing spatial clustering of decreases/increases in the population size of a particular bird (the increase or decrease is the weight, I am using the CONTIGUITY_EDGES_CORNERS Conceptualization of Spatial Relationships). Problematically, I have two polygons with zero nearest neighbors (no overlap, edges/corners touching--think islands). The optimized hotspot analysis runs without any issues, and one of these islands emerges as a significant cold spot. Are these results valid, or should I exclude these polygons from the analysis and only use contiguous polygons? Thank you!

DanPatterson_Retired · Answer

You did read the discussion here.... your call, but if you are looking for spatially contiguous processes and you have 'islands' how contiguous is your spatial sampling frame.  It would come as no surprise that they would be a tad chilly.  I would have a look at what happens when you remove them.  One wouldn't want their attention drawn to something on a map (ie hot or cold) that shouldn't really be there... or would one