I am trying to describe/find statistically spatial patterns in a point data set of industrial sites. The points feature apart from
- their location,
- a date for their opening,
- a date for their closure and
- the length of operation, but no other value.
I would like to see at which moment in time certain patterns emerge and disappear, such as sudden dense cluster or dispersal. I investigated the data set 'manually' and found several moments in time, which show the kind of patterns I am looking for (see image 1). I would however like to explore and describe these statistically, also most likely there are relations through out time, which I wouldn't find else wise.
I am currently working with ArcMap 10.2.2 and have explored some possibilities of space-time analysis. I tried several methods, such as Average Nearest Neighbour, Grouping or Multi-Distance Spatial Cluster Analysis (Ripleys K Function), the difficulty is that these method do not incorporate the time dimension. The best I could find is as described here (http://video.esri.com/watch/1681/spatial-pattern-analysis-mapping-trends-and-clusters) a time space analysis using a Spatial Weight Matrix (Although here I have to introduce a bias in the length of relation [how long and how far away events should be so that the analysis considers them as related]), in combination with a HotSpot analysis, followed by 3 dimensional representation.
The problem I have with this is, that my data set doesn't feature a value or magnitude field against which I could apply a HotSpots analysis. The only values I have are Dates, or respectively time length, which I can not apply a hot spot analysis on.
Does anyone of you maybe has a good idea to analyse such a point data set?