Skip to contents

Draw a Clustered Spatial Sample

Source: R/borg_clustered_sample.R
borg_clustered_sample.Rd

Generates spatially clustered sample points using a parent-offspring process: cluster centres are placed uniformly across the domain, and each sample point is drawn near a randomly assigned centre. Clustered designs are the common ecological sampling pattern and produce the strong spatial dependence that random cross-validation understates, making this useful for simulations, teaching, and stress-testing evaluation schemes.

Usage

borg_clustered_sample(area, n, n_clusters = 5, radius = NULL, seed = NULL)

Arguments

area

The sampling domain. One of: a numeric vector c(xmin, xmax, ymin, ymax); a data frame or matrix of coordinates (its bounding box is used); or an sf polygon or SpatVector polygon, in which case points are constrained to lie inside it.

n

Integer. Total number of sample points to draw.

n_clusters

Integer. Number of cluster centres (parents). Default: 5.

radius

Numeric. Cluster spread: the maximum offset of a point from its centre, in coordinate units. If NULL, set to 5\ diagonal.

seed

Integer or NULL. Random seed for reproducibility. Default: NULL (no seeding).

Value

A data frame with columns x, y, and cluster (the integer id of the assigned centre). Attributes n_clusters and radius record the design. Drops directly into borg_diagnose() and borg_cv() via coords = c("x", "y").

See also

borg_simulate, borg_diagnose

Examples

set.seed(1)
s <- borg_clustered_sample(c(0, 100, 0, 100), n = 120, n_clusters = 6)
nrow(s)
length(unique(s$cluster))