relevant: Extract the minimal sub-Gaussian mixture graphical model required to infer a subset of nodes

Description

This function extracts the minimal sub-Gaussian mixture graphical model required to infer a subset of nodes (i.e. the sub-model relevant to these nodes). The nodes that do not contribute to inference are removed, which includes those that are d-separated from the inferred ones by the nodes whose values are observed, as well as the barren nodes (Druzdzel and Suermondt, 1994).

Usage

relevant(gmgm, nodes, nodes_obs = NULL, nodes_miss = NULL)

Value

The gmbn or gmdbn object relevant to the subset of nodes.

Arguments

gmgm: An object of class gmbn or gmdbn.
nodes: A character vector containing the inferred nodes.
nodes_obs: A character vector containing the nodes whose values are observed.
nodes_miss: A character vector containing the nodes whose values are missing. Note that if a node is neither in nodes_obs nor in nodes_miss, its observability is considered uncertain or varying. Thus, it is not treated as an observed node, nor can it be removed as a barren node.

References

Druzdzel, M. J. and Suermondt, H. J. (1994). Relevance in Probabilistic Models: "Backyards" in a "Small World". In Working Notes of the AAAI 1994 Fall Symposium Series: Relevance, pages 60--63, New Orleans, LA, USA.

Examples

Run this code

data(gmbn_body)
gmbn_1 <- relevant(gmbn_body, "AGE",
                   nodes_obs = c("FAT", "HEIGHT", "WEIGHT"),
                   nodes_miss = "GLYCO")

data(gmdbn_air)
gmdbn_1 <- do.call("gmdbn", gmdbn_air[c("b_1", "b_2")])
gmdbn_2 <- relevant(gmdbn_1, "O3", nodes_obs = "NO2")