inlabru (version 2.1.9)

ipoints: Generate integration points

Description

This function generates points in one or two dimensions with a weight attached to each point. The weighted sum of a function evaluated at these points is the integral of that function approximated by linear basis functions. The parameter region describes the area(s) integrated over.

In case of a single dimension region is supposed to be a two-column matrix where each row describes the start and end point of the interval to integrate over. In the two-dimensional case region can be either a SpatialPolygon, an inla.mesh or a SpatialLinesDataFrame describing the area to integrate over. If a SpatialLineDataFrame is provided it has to have a column called 'weight' in order to indicate the width of the line.

The domain parameter is an inla.mesh.1d or inla.mesh object that can be employed to project the integration points to the vertices of the mesh. This reduces the final number of integration points and reduces the computational cost of the integration. The projection can also prevent numerical issues in spatial LGCP models where each observed point is ideally surrounded by three integration point sitting at the coresponding mesh vertices. For convenience, the domain parameter can also be a single integer setting the number of equally spaced integration points in the one-dimensional case.

Usage

ipoints(region = NULL, domain = NULL, name = "x", group = NULL, project)

Arguments

region

Description of the integration region boundary. In 1D either a vector of two numerics or a two-column matrix where each row describes and interval. In 2D either a SpatialPolygon or a SpatialLinesDataFrame with a weight column defining the width of the line.

domain

In 1D a single numeric setting the numer of integration points or an inla.mesh.1d defining the locations to project the integration points to. In 2D domain has to be an inla.mesh object describing the projection and granularity of the integration.

name

Character array stating the name of the domains dimension(s)

group

Column names of the region object (if applicable) for which the integration points are calculated independently and not merged by the projection.

project

If TRUE, project the integration points to mesh vertices

Value

A data.frame or SpatialPointsDataFrame of 1D and 2D integration points, respectively.

Examples

# NOT RUN {
if (require("INLA", quietly = TRUE)) {

# Create 50 integration points covering the dimension 'myDim' between 0 and 10. 

ips = ipoints(c(0,10), 50, name = "myDim")
plot(ips)

# Create integration points for the two intervals [0,3] and [5,10]

ips = ipoints(matrix(c(0,3, 5,10), nrow = 2, byrow = TRUE), 50)
plot(ips)

# Convert a 1D mesh into integration points
mesh = inla.mesh.1d(seq(0,10,by = 1))
ips = ipoints(mesh, name = "time")
plot(ips)


# Obtain 2D integration points from a SpatialPolygon

data(gorillas, package = "inlabru")
ips = ipoints(gorillas$boundary)
ggplot() + gg(gorillas$boundary) + gg(ips, aes(size = weight))


#' Project integration points to mesh vertices

ips = ipoints(gorillas$boundary, domain = gorillas$mesh)
ggplot() + gg(gorillas$mesh) +  gg(gorillas$boundary) + gg(ips, aes(size = weight))


# Turn a 2D mesh into integration points

ips = ipoints(gorillas$mesh)
ggplot() + gg(gorillas$boundary) + gg(ips, aes(size = weight))
}
# }