kernel_smooth: Produce a smooth surface using 2D kernel density smoothing

In Michael-Stevens-27/silverblaze: Geographic Profiling in R

Description

Takes lon/lat coordinates, bins in two dimensions and smooths using kernel density smoothing. Kernel densities are computed using the fast Fourier transform method, which is many times faster than simple summation when using a large number of points. Each Kernel is student's-t distributed and scaled by the bandwidth lambda. If lambda is set to NULL then the optimal value of lambda is chosen automatically using the leave-one-out maximum likelihood method.

Usage

kernel_smooth(
  longitude,
  latitude,
  breaks_lon,
  breaks_lat,
  lambda = NULL,
  nu = 3
)

Arguments

longitude	longitude of input points
latitude	latitude of input points
breaks_lon	positions of longitude breaks
breaks_lat	positions of latitude breaks
lambda	bandwidth to use in posterior smoothing. If NULL then optimal bandwidth is chosen automatically by maximum-likelihood
nu	degrees of freedom of student's-t kernel

References

Barnard, Etienne. "Maximum leave-one-out likelihood for kernel density estimation." Proceedings of the Twenty-First Annual Symposium of the Pattern Recognition Association of South Africa. 2010

Michael-Stevens-27/silverblaze documentation built on May 28, 2021, 5:47 p.m.