Skip to contents

Group PUMAs by sample rate using spectral clustering

spectral_zone_groups.Rd

Create PUMA zone groups with similar sample rates and adjacency using spectral clustering. Use to balance sample rates and spatial contiguity for weighting pipelines.

Usage

spectral_zone_groups(
  seed,
  targets,
  puma_sf,
  k_groups = "auto",
  rng_seed = 4321
)

Arguments

seed

data.table with required columns:

  • puma_id — PUMA identifier

  • p_total — total persons

  • h_total — total households Rows: one per household. Keys: (puma_id). Modified by reference: no (returns copy).

targets

data.table with required columns:

  • puma_id

  • p_total

  • h_total Rows: one per household. Keys: (puma_id). Modified by reference: no.

puma_sf

sf object. PUMA geometries. Must include PUMA ID column.

k_groups

integer or 'auto'. Number of groups to create. If 'auto', selects optimal k.

rng_seed

integer. Random seed for reproducibility. Default 4321.

Value

data.table with columns:

  • puma_id — PUMA identifier

  • zone_group — assigned group Rows: one per PUMA. Keys: (puma_id).

Details

  • Constructs adjacency matrix from PUMA geometries (sf polygons).

  • Calculates node and edge weights using household sample rates.

  • Computes Laplacian matrix and eigenvectors for clustering.

  • Assigns zone groups via k-means on spectral features.

  • Returns a copy; does not modify by reference.

  • Assumes valid PUMA IDs and geometries; errors if missing.

Settings

None.

See also

  • scripts/weighting/zone_groups.R

Other clustering utilities: entropy_zone_groups(), get_kmeans_opt_k(), kmeans_zone_groups()

Examples

## Not run:
spectral_zone_groups(seed, targets, puma_sf, k_groups = 3)
#> Error: object 'seed' not found
## End(Not run)