[R] 2D density tophat

Aaron Robotham asgr at st-andrews.ac.uk
Fri Dec 5 16:51:13 CET 2008


In case anyone other than me was interested, a pretty efficient  
circular tophat can be made using the fields function fields.rdist.near:

CircHat=function (x, y, h=1, gridres = c((max(x)-min(x))/25,(max(y)- 
min(y))/25), lims = c(range(x), range(y)),density=FALSE)
{
	require(fields)
     nx <- length(x)
     ny <- length(y)
     n=c(1+(lims[2]-lims[1])/gridres[1],1+(lims[4]-lims[3])/gridres[2])
     if (length(y) != nx)
         stop("data vectors must be the same length")
     if (any(!is.finite(x)) || any(!is.finite(y)))
         stop("missing or infinite values in the data are not allowed")
     if (any(!is.finite(lims)))
         stop("only finite values are allowed in 'lims'")
     gx <- seq(lims[1], lims[2], by = gridres[1])
     gy <- seq(lims[3], lims[4], by = gridres[2])
     fullgrid=expand.grid(gx,gy)

     if (missing(h))
         h <- c(bandwidth.nrd(x), bandwidth.nrd(y))
     
temp 
= 
table 
(fields 
.rdist 
.near 
(as 
.matrix 
(fullgrid 
),as.matrix(cbind(x,y)),mean.neighbor=ceiling(length(x)*pi*h^2/ 
((lims[2]-lims[1])*(lims[4]-lims[3]))),delta=h)$ind[,1])
     	
pad=rep(0,length(gx)*length(gy))
pad[as.numeric(names(temp))]=as.numeric(temp)

z <- matrix(pad, length(gx), length(gy))

if(density){z=z/(nx*pi*h^2)}

return=list(x = gx, y = gy, z = z)

}

It works in more or less the same way as kde2d, but by default it  
returns counts not densities

Aaron

On 1 Dec 2008, at  11:46, Aaron Robotham wrote:

> Hello R users,
>
> I have successfully created a square (or more generally, rectangular)
> tophat smoothing routine based on altering the already available
> KDE2D. I would be keen to implement a circular tophat routine also,
> however this appears to be much more difficult to write efficiently (I
> have a routine, but it's very slow). I tried to create one based on
> using crossdist (in spatstat) to create a distance matrix between my
> data and the sampling grid, but it doesn't take a particularly large
> amount of data (or hi-res grid) for memory to be a big problem. The 2D
> density routines I have been able to find either don't support a
> simple tophat, or don't use the absolute distances between the
> sampling grid and the data. Should anyone know of more general 2D
> density routines that might support circular tophats, or know of a
> simple and efficient method for creating them, I would be very  
> grateful.
>
> Thanks for your time,
>
> Aaron
>
> PS: I tried sending this on Friday originally, but as far as I know
> that didn't work, so should another post appear from me asking the
> same thing I apologise in advance.
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.



More information about the R-help mailing list