synergistic-unique-redundant decomposition of causality

Usage

# S4 method for class 'data.frame'
surd(
  data,
  target,
  agents,
  lag = 1,
  bin = 5,
  max.combs = NULL,
  cores = 1,
  backend = "threading"
)

# S4 method for class 'sf'
surd(
  data,
  target,
  agents,
  lag = 1,
  bin = 5,
  max.combs = NULL,
  cores = 1,
  backend = "threading",
  nb = NULL
)

# S4 method for class 'SpatRaster'
surd(
  data,
  target,
  agents,
  lag = 1,
  bin = 5,
  max.combs = NULL,
  cores = 1,
  backend = "threading"
)

Arguments

data: observation data.
target: name of the target variable.
agents: names of agent variables.
lag: (optional) lag order.
bin: (optional) number of discretization bins.
max.combs: (optional) maximum combination order. If NULL, the standard SURD decomposition is applied.
cores: (optional) number of cores for parallel computation.
backend: (optional) Joblib backend: loky, threading, or multiprocessing.
nb: (optional) neighbours list.

Value

A list.

unique: Unique information contributions per variable.
synergistic: Synergistic information components by agent combinations.
redundant: Redundant information shared by agent subsets.
mutual_info: Mutual information measures for each combination.
info_leak: Information leak ratio.

References

Martínez-Sánchez, Á., Arranz, G. & Lozano-Durán, A. Decomposing causality into its synergistic, unique, and redundant components. Nat Commun 15, 9296 (2024).

Examples

columbus = sf::read_sf(system.file("case/columbus.gpkg", package="spEDM"))
# \donttest{
tryCatch(
  surd(columbus, "hoval", c("inc", "crime")),
  error = \(e) message("Skipping Python-dependent example: ", e$message)
)
# }