Skip to content

spatialpoints: each dot represents 100 individuals?

6 messages · Juta Kawalerowicz, Santiago Beguería, Tim Appelhans +2 more

Juta Kawalerowicz
# 
Hi,

I have a dataset with couple of million of points (individuals) and
would like to do some mapping (I have the coordinates of each point)
but given the number of observation I think it may be usuful to plot
dots which represent 100 individuals (of a given group). Does anyone
know a good way to aggregate up spatialpoints? Any suggestions would
be much appreciated!

Best wishes,
Juta
Santiago Beguería
# 
Hi Juta,

Do you need to represent the aggregated data as points necessarily? Otherwise, I have used the stat_hexbin representation from ggplot2 for that very purpose, see for instance:

http://santiago.begueria.es/wordpress/wp-content/uploads/2015/06/Figure_01.png

Cheers,

Stg
Tim Appelhans
# 
Hi Juta,
in case you want to do web-mapping, library("mapview") should be able to 
handle a couple of million (in the sense that a couple means approx. 2).

Cheers
Tim
On 25.04.2016 13:35, Juta Kawalerowicz wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Edzer Pebesma
    

    

      
      #
    

  


  

      
Do these hexagons represent areas with equal area?

        
On 25/04/16 13:52, Santiago Beguer?a wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Santiago Beguería
    

    

      
      #
    

  


  

      
Good question. Not on my example, I don?t think so. It will depend on the projection of your data, I guess.

Stg

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    
2 days later

    

      
      

        


  

        

      rubenfcasal
    

    

      
      #
    

  


  

      
Alternatively, you might also consider data binning (implemented in 
several packages: KernSmooth, ks, sm, npsp ,...). This technique is 
commonly used in nonparametric statistics to reduce the computational 
time (see e.g. Wand, M. P. (1994), Fast Computation of Multivariate 
Kernel Estimators, Journal of Computational and Graphical Statistics, 3, 
433-445).

For instance, using the npsp package (maintained by me...), you could do 
something like this:

library(npsp)

bin <- binning(earthquakes[, c("lon", "lat")], nbin = c(50,50))

# ?bin$binw? will contain the binning weights (aggregations) at 
locations ?coords(bin)?

simage(bin)

Additionally, you could estimate (nonparametrically) the spatial density:

h <- h.cv(bin, ncv = 2)$h

den <- np.den(bin, h = h)

plot(den, log = FALSE, main = 'Estimated density')

Best regards,

Ruben.


El 25/04/2016 a las 13:35, Juta Kawalerowicz escribi?: