strength.plot: Arc strength plot

In bnlearn: Bayesian Network Structure Learning, Parameter Learning and Inference

Arc strength plot

Description

Plot a Bayesian network and format its arcs according to the strength of the dependencies they represent. Requires the Rgraphviz package.

Usage

strength.plot(x, strength, threshold, cutpoints, highlight = NULL, groups,
  layout = "dot", shape = "rectangle", fontsize = 12, main = NULL, sub = NULL,
  render = TRUE, debug = FALSE)

Arguments

x	an object of class bn.
strength	an object of class bn.strength computed from the object of class bn corresponding to the x argument.
threshold	a numeric value. See below.
cutpoints	an array of numeric values. See below.
highlight	a list, see graphviz.plot for details.
groups	a list of character vectors, representing groups of node labels of nodes that should be plotted close to each other.
layout	a character string, the layout argument that will be passed to Rgraphviz. Possible values are dots, neato, twopi, circo and fdp. See Rgraphviz documentation for details.
shape	a character string, the shape of the nodes. Can be circle, ellipse or rectangle.
fontsize	a positive number, the font size for the node labels.
main	a character string, the main title of the graph. It's plotted at the top of the graph.
sub	a character string, a subtitle which is plotted at the bottom of the graph.
debug	a boolean value. If TRUE a lot of debugging output is printed; otherwise the function is completely silent.
render	a logical value. If TRUE, strength.plot() actually draws the figure in addition to returning the corresponding graph object. If FALSE, no figure is produced.

Details

The threshold argument is used to determine which arcs are supported strongly enough by the data to be deemed significant:

• if arc strengths have been computed using conditional independence tests, any strength coefficient (which is the p-value of the test) lesser than or equal to the threshold is considered significant.

• if arc strengths have been computed using network scores, any strength coefficient (which is the increase/decrease of the network score caused by the removal of the arc) lesser than the threshold is considered significant.

• if arc strengths have been computed using bootstrap or using Bayes factors, any strength coefficient (which can be interpreted as a probability for inclusion) greater or equal than the threshold is considered significant.

The default value is the value of the strength attribute of the bn.strength object passed via the strength argument.

Non-significant arcs are plotted as dashed lines.

The cutpoints argument is an array of numeric values used to divide the range of the strength coefficients into intervals. The interval each strength coefficient falls into determines the line width of the corresponding arc in the plot. The default intervals are delimited by

unique(c(0, threshold/c(10, 5, 2, 1.5, 1), 1))

if the coefficients are computed from conditional independence tests, by

unique(c(0, (1 - threshold)/c(10, 5, 2, 1.5, 1), 1))

for bootstrap estimates or by the quantiles

quantile(s[s < threshold], 1 - c(0.50, 0.75, 0.90, 0.95, 1))

of the significant differences if network scores are used.

Value

graphviz.plot() returns invisibly the graph object produced by Rgraphviz. It can be further modified using the commands present in the graph and Rgraphviz packages, and it contains the arc strengths in the edge weight attribute.

Author(s)

Marco Scutari

Examples

## Not run: 
# plot the network learned by hc().
dag = hc(learning.test)
strength = arc.strength(dag, learning.test, criterion = "x2")
strength.plot(dag, strength)
# add another (non-significant) arc and plot the network again.
dag = set.arc(dag, "A", "C")
strength = arc.strength(dag, learning.test, criterion = "x2")
strength.plot(dag, strength)

## End(Not run)

bnlearn documentation built on Sept. 11, 2024, 8:27 p.m.