Skip to content

Distances between two datasets of x and y co-ordinates

2 messages · Adrian Baddeley, Liaw, Andy

Adrian Baddeley
# 
Andrew McFadden <Andrew.McFadden at maf.govt.nz> writes:
This can be done efficiently in the package 'spatstat'.

     library(spatstat)
     crossdist(x1, y1, x2, y2)

where x1, y1 are vectors of coordinates for the first set of points
and x2, y2 for the second set. See help(crossdist.default)

This is executed in C and is faster than using outer() or apply().

The result is a matrix giving the distance between each pair of points
(the first point in the first dataset and the second point in the second
set). If these datasets are large, you can of course run into trouble with
the size of this matrix.

If you just wanted to find the distance to the *nearest* point
(or identify which point is nearest), use the function nncross().

Adrian Baddeley
Liaw, Andy
# 
From: adrian at maths.uwa.edu.au
The margin may not be as large as you imagine:

crossdist2 <- function(x1, y1, x2, y2) {
    c1 <- complex(real=x1, imaginary=y1)
    c2 <- complex(real=x2, imaginary=y2)
    dist <- outer(c1, c2, function(z1, z2) Mod(z1-z2))
    dist
}

R> set.seed(17)
R> n <- 1000
R> x1 <- rnorm(n)
R> y1 <- rnorm(n)
R> x2 <- rnorm(n)
R> y2 <- rnorm(n)
R> system.time(d2 <- crossdist2(x1, y1, x2, y2))
   user  system elapsed 
   1.03    0.09    1.17 
R> library(spatstat)
R> system.time(d1 <- crossdist2(x1, y1, x2, y2))
   user  system elapsed 
   1.08    0.03    1.12 
R> all.equal(d1, d2)
[1] TRUE

Andy
------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachme...{{dropped:15}}