[postgis-users] Creating density map

[Ben Madin]

Aren,

You might have more luck on the R-sig-geo list. This looks like a mix of network analysis and point processes - 

check out pgRouting, or look for routing algorithms
check out the sp package and then spatstat, network etc.

An approach (by no means the right or even a good one) is to think of it as a graph or MCMC problem, and consider the relationship between the events as probabilities that can be affected by distance. You are of course trying to create a complex spatio-temporal model like everyone, so there is a fair bit of literature out there. It probably pays to look at relevant articles in your field. 

Once you have this clearer, I think you should be able to work out how to get the data into R - but it could either be as a data frame using spatial SQL to return attributes, or using readOGR if you need spatial objects in R

cheers

Ben

 


On 10/01/2011, at 2:50 PM, Aren Cambre wrote:

> I have three datasets:
> Routes
> Event type A that occurs along the routes (points)
> Event type B that occurs along the routes (points)
> Both event types have several attributes, including a date/timestamp, sub-classes of each event type, and other meaningful attributes.
> 
> I'm trying to use statistical methods to check for certain relationships between event types A and B. They may influence each other (A may affect B and B may affect A). I also want to see if there's a relationship between subtypes. E.g., do events A.X or A.Y have a stronger impact on event type B?
> 
> I'd like to make heat density maps to help interpret the data, but I have two conceptual problems.
> 
> First problem is how to make the map. The programmatically easy but slow way is to create a greedy algorithm to traverse every route. During traversal, create a point at each increment of distance X. An attribute of that point may be the number of qualifying events no more than distance Y from that point.
> 
> I may need to limit to events along the route I am traversing. E.g., if traversing route M looking for event type B, and I come across route N, the heat map for route M probably should not include events of type B along route N event if they are within Y distance from my current point.
> 
> Second problem is how to deal with all the permutations. I could muck through the simple algorithm and make spiffy point maps, and with a little graphical wizardry, I could make the maps pretty. However, I need to do analysis over different time periods. E.g., does the relative intensity of week I's event type As along route M affect the occurrence of event type B on week I+1? How about event type A.X? A.Y? Do they have different effects over the same time period? I have between 3 and 9 years of event types A and B...
> 
> All the permutations (not simply combinations) of factors can really explode the complexity of this project.
> 
> To prevent wheel reinventing, are there already well-tread solutions to this problem? I've done some Google searches and am not coming up with much, so I guess I may not be using the correct lingo?
> 
> I know that I need to incorporate R into this at some point; my objective now is to get the data to a point where I could use R to analyze it.
> 
> Aren
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20110110/eb260c27/attachment.html>