Plotting
You can visualize causal graphs using the plot()
function, which works for both knowledge and
knowledgeable_caugi objects. The function leverages the
underlying caugi::plot() method, providing flexible options
for customizing the appearance of nodes and edges.
Plotting Knowledge Objects
We start by creating a simple Knowledge object and
visualizing it with plot(). This knowledge encodes the
following relationships between variables in our dataset:
data(num_data)
kn <- knowledge(
num_data,
X1 %-->% c(X2, X3), # Require edge from X1 to X2, and X1 to X3
X2 %!-->% c(X3, Y), # Forbid edge from X2 to X3, and X2 to Y
Y %!-->% Z # Forbid edge from Y to Z
)
plot(kn)
By default, required edges are shown in blue and forbidden edges in
red. You can customize these colors using the required_col
and forbidden_col arguments:
Different node layouts, node styles, and edge styles can be specified
using the layout, node_style, and
edge_style parameters. We provide a quick example and refer
to the causalDisco::plot() function (and the underlying
caugi::plot() function) documentation for more details.
Here we use the "fruchterman-reingold" layout and customize
node and edge styles:
plot(
kn,
layout = "fruchterman-reingold",
node_style = list(
fill = "lightblue", # Fill color
col = "darkblue", # Border color
lwd = 2, # Border width
padding = 4, # Text padding (mm)
size = 1.2 # Size multiplier
),
edge_style = list(
lwd = 1.5, # Edge width
arrow_size = 4 # Arrow size (mm)
),
required_col = "blue", # Color for required edges
forbidden_col = "red" # Color for forbidden edges
)
Note that to override the colors for required and forbidden edges
only for a specific edge, the edge needs to be targeted explicitly in
the by_edge list of the edge_style argument.
For example, to make the required edge from X1 to
X2 orange and the forbidden edge from X2 to
Y yellow, while letting all other edges use the default
colors, we can do:
plot(
kn,
layout = "fruchterman-reingold",
node_style = list(
fill = "lightblue", # Fill color
col = "darkblue", # Border color
lwd = 2, # Border width
padding = 4, # Text padding (mm)
size = 1.2 # Size multiplier
),
edge_style = list(
lwd = 1.5, # Edge width
arrow_size = 4, # Arrow size (mm)
# Per-edge overrides
by_edge = list(
X1 = list(
X2 = list(col = "orange", fill = "orange", lwd = 3)
),
X2 = list(
Y = list(col = "yellow", fill = "yellow", lwd = 3)
)
)
),
required_col = "blue", # Color for required edges
forbidden_col = "red" # Color for forbidden edges
)
You can also pass a custom layout to plot():
my_layout <- data.frame(
name = c("X1", "X2", "X3", "Y", "Z"),
x = c(1, 2, 3, 4, 5),
y = c(1, 2, 1, 2, 1)
)
plot(kn, layout = my_layout)
Plotting Tiered Knowledge
We can also visualize tiered knowledge structures. Here is an example using a dataset with variables measured at different life stages:
data(tpc_example)
kn_tiered <- knowledge(
tpc_example,
tier(
child ~ starts_with("child"), # tidyselect helper; equivalent to c("child_x1", "child_x2")
youth ~ starts_with("youth"),
old ~ starts_with("old")
),
child_x1 %-->% child_x2,
child_x2 %!-->% youth_x3
)
plot(kn_tiered)
Variables are grouped by life stage, reflecting the tiered structure. This helps visually convey temporal or hierarchical relationships in the data.
Plotting Knowledgeable Caugi Objects
Here we run a causal discovery algorithm to get a Disco
object, which is the knowledge object alongside the learned causal graph
as a caugi object. We can then plot this combined object to
see both the prior knowledge and the learned relationships
data(num_data)
kn <- knowledge(
num_data,
X1 %-->% X2,
X2 %!-->% c(X3, Y),
Y %!-->% Z
)
pc_bnlearn <- pc(engine = "bnlearn", test = "fisher_z")
pc_result <- disco(num_data, method = pc_bnlearn, knowledge = kn)
plot(pc_result)
Blue edges indicate required relationships from prior knowledge, black edges show those learned from the data, and forbidden edges are not plotted (to avoid confusion and clutter).
With tiered knowledge, the plotting works similarly:
data(tpc_example)
kn_tiered <- knowledge(
tpc_example,
tier(
child ~ starts_with("child"),
youth ~ starts_with("youth"),
old ~ starts_with("old")
)
)
cd_tges <- tges(engine = "causalDisco", score = "tbic")
disco_cd_tges <- disco(data = tpc_example, method = cd_tges, knowledge = kn_tiered)
plot(disco_cd_tges)
Where we can see the tiers reflected in the layout of the graph.
