interpolateAndApplyWithinSpatial: Generic function to interpolate from a polygon to points...
In gsk3/taRifx.geo: Collection of various spatial functions

Description Usage Arguments Value Examples

Generic function to interpolate from a polygon to points lying inside it This function was designed to solve the following problem. Suppose you have counts of the number of entities inside a polygon (N). To compute distances to a point, you might take the distance from the polygon centroid. But this is too simplistic–it discards the positional uncertainty inherent in not knowing the exact location of each entity which makes up the count. Instead, we repeatedly sample N points from the census block group centroids which lie within our polygon, weight them by their population, and compute distances from there.

1 2	interpolateAndApplyWithinSpatial(crude, fine, FUN, nSampleCol, samplesize = 30, simplify = FALSE, ...)

`crude`	A SpatialPolygonsDataFrame containing the variable of interest
`fine`	an sp object whose points or polygons lie inside the polygons of crude
`FUN`	An function taking at least two arguments (a single polygon of crude in the first position, and the within-polygon elements of fine in the second)
`nSampleCol`	Column name in crude containing number of elements of fineWithin to sample per polygon
`samplesize`	The number of replicates per element of crude to draw
`simplify`	Whether to simplify to an array or not
`...`	Arguments to be passed to FUN

list of length length(crude) where each element is a list of length samplesize containing the results of FUN for that crude-element/sample

# Not run because too time-consuming
## Not run: 
require(fields)
require(rgdal)
distanceMatrix <- function( points1, points2, dist.fn=rdist.earth ) {
 cat( "Generating distance matrix for ",length(points1)," by ", length(points2), " matrix.\n" )
 if(!is.na(proj4string(points1))) {
   points1 <- sp::spTransform( points1, sp::CRS("+proj=longlat +datum=WGS84") )
 }
 if(!is.na(proj4string(points2))) {
   points2 <- sp::spTransform( points2, sp::CRS("+proj=longlat +datum=WGS84") )
 }
 dist.fn( coordinates(points1), coordinates(points2) )
}
# One option: Use the apply functionality
dist <- interpolateAndApplyWithinSpatial(
 crude=polySP, fine=pointSP,
 FUN=distanceMatrix,
 nSampleCol="z", samplesize=25,
 points2=pointSP2, simplify=TRUE
)
# Dist now is a list of 3 matrices, each with dimensions: N x length(pointSP2) x samplesize
# Each matrix represents N entities imputed from a single polygon,
#   so we can actually simplify further
library(abind)
distmat <- do.call( Curry(abind, along=1), dist )
mindist <- apply( distmat, 3, function(x) { # For each realization of the 'world'
  apply( x,  )
} )

## End(Not run) # end of dontrun