Hotspot (ArcGis Pro) and cluster analysis (SatScan) for incidents on a transport line: I am confused!

06-05-2019 03:50 AM
New Contributor

Hi all,

Sorry for the long post and the many questions

I am experimenting for my PhD with ArcGIS Pro (I am a psychologist and not a geographer) and I conducted hotspot analysis and a cluster time-space analysis on SatScan (I don't think I have ever found a program so difficult to use in my life!!!). I have point data with time stamps over 4 years, and I want to see if there are hotspots. Because the type of incident is rare, every point comes up as a hotspot. I was wondering if there is a better way to do it?


I want to see if there is an increase, decrease or if there is a pattern (eg if one happens there's more happening at the same location or nearby) and also if we can "profile" the individuals in the incident (I have age, ethnicity and whether it resulted in a fatality outcome or not). Another idea I had would be looking at the number of people passing through that location (it is a transport line) and perhaps do some sort of analysis to show if a higher number of passing people increases the risks (likely) but also where in this transport line there is a low number of passing people and a high incident ratio? Is it possible? I have read Esri's manuals and I can't find an answer! I know time cube is supposed to be the tool to use, but it does not give me hotspots results?

Also, I assume the data is autocorrelated because it is only on that transport line, so Is there something that needs to be adjusted for?


(done on a different dataset where I aggregated incidents per population) I have done the SatScan analysis but I am not quite sure I get it. For example, I ran the analysis for 1 week, 1 months and 1 year but it gives me different locations of clusters time-space. How is that possible? If it is a cluster at week 1, would it not be a cluster at year too? Also, when I am asked to input the threshold in which it checks for risks in the population, what is a correct value? I am read some papers but they all seem to approach it differently.

0 Kudos
0 Replies