clean.graph: cleanup dependence structure graph

In multivariance: Measuring Multivariate Dependence Using Distance Multivariance

Description

Given a dependence structure graph: vertices representing the multivariances of only two vertices can be turned into an edge labeled with the label of the vertex. Moreover, only subsets of the graph can be selected.

Usage

clean.graph(
  g,
  only.level = NULL,
  simplify.pairs = TRUE,
  drop.label.pairs = FALSE
)

Arguments

g	graph, created by dependence.structure
only.level	integer vector, if provided all edges and dependency nodes corresponding to dependence orders not given in 'only.level' are removed
simplify.pairs	boolean, if true dependency nodes which are only connected to two variables are turned into edges
drop.label.pairs	boolean, if true the labels for edges indicating pairwise dependence are removed

Details

Note: The option 'only.level' works only properly for a full dependence structure graph, in the case of a clustered dependence structure graph dependency nodes representing a cluster might be removed.

Value

graph

Examples

N = 200
y = coins(N,2)
x = cbind(y,y,y)

ds = dependence.structure(x,structure.type = "clustered")
plot(clean.graph(ds$graph))
plot(clean.graph(ds$graph,only.level = 2))
plot(clean.graph(ds$graph,only.level = 3)) # of limited use for a clustered graph,
# i.e., here the three-dependence node without edges indicates that
# all edges were connected to clusters

ds = dependence.structure(x,structure.type = "full")
plot(clean.graph(ds$graph))
plot(clean.graph(ds$graph,drop.label.pairs = TRUE))
plot(clean.graph(ds$graph,only.level = 2))
plot(clean.graph(ds$graph,only.level = 2,drop.label.pairs = TRUE))
plot(clean.graph(ds$graph,only.level = 3))

multivariance documentation built on Oct. 6, 2021, 5:08 p.m.