Conditional Mutual Information

Estimate the conditional mutual information between target and interacting variables given conditioning variables.

Usage

cmi(
  data,
  target,
  interact,
  conds,
  base = exp(1),
  type = c("cont", "disc"),
  k = 3,
  normalize = FALSE
)

data: Observation data.
target: Integer vector of column indices for the target variables.
interact: Integer vector of column indices for the interacting variables.
conds: Integer vector of column indices for the conditioning variables.
base: (optional) Logarithm base of the entropy. Defaults to exp(1) (nats). Use 2 for bits or 10 for dits.
type: (optional) Estimation method: "disc" for discrete entropy or "cont" for continuous entropy (KSG estimator).
k: (optional) Number of nearest neighbors used by the continuous estimator. Ignored when type = "disc".
normalize: (optional) Logical; if TRUE, return normalized mutual information.

A numerical value.

set.seed(42)
infoxtr::cmi(matrix(stats::rnorm(99,1,10),ncol=3),1,2,3)
#> [1] 0.02440107