caugi (version 1.1.0)

minimal_d_separator: Compute a minimal d-separator

Description

Computes a minimal d-separator Z for sets X and Y in a DAG, optionally with mandatory inclusions and restrictions on the separator.

Usage

minimal_d_separator(
  cg,
  X = NULL,
  Y = NULL,
  I = character(0),
  R = NULL,
  X_index = NULL,
  Y_index = NULL,
  I_index = NULL,
  R_index = NULL
)

Value

A character vector of node names representing the minimal separator, or NULL if no valid separator exists within the restriction R.

Arguments

cg

A caugi object (must be a DAG).

X, Y

Character vectors of node names. Use *_index to pass 1-based indices.

I

Nodes that must be included in the separator.

R

Nodes allowed in the separator. If NULL, uses all nodes excluding X and Y.

X_index, Y_index, I_index, R_index

Optional numeric 1-based indices (exclusive with corresponding name parameters).

Details

A d-separator Z for X and Y is a set of nodes such that conditioning on Z d-separates X from Y in the graph. This function returns a minimal separator, meaning no proper subset of Z still d-separates X and Y.

The algorithm:

1. Restricts to ancestors of X U Y U I

2. Computes initial separator candidate from R

3. Refines using Bayes-ball d-connection algorithm

4. Returns minimal separator or NULL if none exists within R

See Also

Other adjustment: adjustment_set(), all_adjustment_sets_admg(), all_backdoor_sets(), d_separated(), is_valid_adjustment_admg(), is_valid_backdoor()

Examples

cg <- caugi(
  A %-->% X,
  X %-->% M,
  M %-->% Y,
  A %-->% Y,
  class = "DAG"
)

# Find any minimal separator between X and Y
minimal_d_separator(cg, "X", "Y")

# Force M to be in the separator
minimal_d_separator(cg, "X", "Y", I = "M")

# Restrict separator to only {A, M}
minimal_d_separator(cg, "X", "Y", R = c("A", "M"))