Skip to contents

Discretise a bivariate Gaussian kernel on a grid

Source: R/calibrar-splines.R
gaussian_kernel.Rd

Evaluate a two-dimensional Gaussian density on a regular grid defined by lower and upper bounds in each dimension.

Usage

gaussian_kernel(par, lower, upper, n = 10, checkSymmetry = TRUE, ...)

Arguments

par

A list containing at least mean and sigma, where mean is a numeric vector of length two and sigma is a 2 x 2 covariance matrix.

lower

Numeric vector of length two giving the lower bounds of the grid.

upper

Numeric vector of length two giving the upper bounds of the grid.

n

Number of grid cells per dimension. This can be a scalar or a length-two numeric vector.

checkSymmetry

Logical. Passed to mvtnorm::dmvnorm() when checking the covariance matrix.

...

Further arguments reserved for future extensions.

Value

A list with components x, y, and z, suitable for contour or image plotting. x and y contain the grid coordinates, and z contains the corresponding matrix of Gaussian density values.

Details

This helper is convenient when a smooth bivariate dispersal or movement kernel needs to be represented on a finite grid. The function uses mvtnorm::dmvnorm() internally and returns the density values without normalising them to sum to one over the discretised domain.

See also

mvtnorm::dmvnorm()

Examples

ker = gaussian_kernel(
  par = list(mean = c(0, 0), sigma = diag(2)),
  lower = c(-2, -2),
  upper = c(2, 2),
  n = c(10, 10)
)
#> Error in do.call(expand.grid, lim): second argument must be a list
str(ker)
#> Error: object 'ker' not found