hotbox {spatstat.explore}

Heat Kernel for a Two-Dimensional Rectangle

Description

Calculate values of the heat kernel in a rectangle with insulated edges.

Usage

hotbox(Xsource, Xquery, sigma,
       ..., W=NULL, squared=FALSE, nmax=20)

Arguments

Xsource

Point pattern of sources of heat. Object of class "ppp" or convertible to a point pattern using as.ppp(Xsource, W).

Xquery

Locations where the heat kernel value is required. An object of class "ppp" specifying query location points, or an object of class "im" or "owin" specifying a grid of query points.

sigma

Bandwidth for kernel. A single number.

...

Extra arguments (passed to as.mask) controlling the pixel resolution of the result, when Xquery is a window or an image.

W

Window (object of class "owin") used to define the spatial domain when Xsource is not of class "ppp".

squared

Logical value indicating whether to take the square of each heat kernel value, before summing over the source points.

nmax

Number of terms to be used from the infinite-sum expression for the heat kernel. A single integer.

Details

This function computes the sum of heat kernels associated with each of the source points, evaluating them at each query location.

The window for evaluation of the heat kernel must be a rectangle.

The heat kernel in any region can be expressed as an infinite sum of terms associated with the eigenfunctions of the Laplacian. The heat kernel in a rectangle is the product of heat kernels for one-dimensional intervals on the horizontal and vertical axes. This function uses hotrod to compute the one-dimensional heat kernels, truncating the infinite sum to the first nmax terms, and then calculates the two-dimensional heat kernel from each source point to each query location. If squared=TRUE these values are squared. Finally the values are summed over all source points to obtain a single value for each query location.

Value

If Xquery is a point pattern, the result is a numeric vector with one entry for each query point.

If Xquery is an image or window, the result is a pixel image.

Author(s)

Adrian Baddeley and Greg McSwiggan.

References

Baddeley, A., Davies, T., Rakshit, S., Nair, G. and McSwiggan, G. (2021) Diffusion smoothing for spatial point patterns. Statistical Science, in press.

See Also

densityHeat.ppp

Examples

  X <- runifpoint(10)

  Y <- runifpoint(5)
  hotbox(X, Y, 0.1)

  plot(hotbox(X, Window(X), 0.1))
  points(X, pch=16)