Visualizing graphs in CFID

Plotting with dagitty

Plotting the graph by exporting a dagitty graph and using ggdag. It is best not to load dagitty directly, as it’s namespace clashes with cfid. Dagitty export doesn’t work with parallel graphs Dagitty doesn’t accept the special characters that cfid uses in labeling variables in counterfactual graphs and parallel world graph.

g <- cfid::dag("X -> W -> Y <- Z <- D X <-> Y")
dagitty_g <- export_graph(g, "dagitty")
ggdag::ggdag(dagitty_g) + ggdag::theme_dag_gray()

Plotting with bnlearn and rgraphviz

The Bayesian network package bnlearn has a good interface with graphviz. Plotting this way requires less boilerplate then using rgraphviz directly. Rgraphviz needs to be installed via bioconductor.

plot_graphviz <- function(g){
  el <- get_edgelist(g)
  nodes_ <- unique(unlist(el))
  net <- bnlearn::empty.graph(nodes=nodes_)
  bnlearn::arcs(net) <- el
  bnlearn::graphviz.plot(net)
}
plot_graphviz(g)
#> Loading required namespace: Rgraphviz

Plotting a parallel world graph.

v1 <- cf("Y", 0, c(X = 0))
v2 <- cf("X", 1)
v3 <- cf("Z", 0, c(D = 0))
v4 <- cf("D", 0)
gamma <- conj(v4, v1, v2, v3)
pw_graph <- cfid:::pwg(g, gamma)
plot_graphviz(pw_graph)

Plotting a counterfactual graph

cf_graph <- cfid:::make_cg(g, gamma)
plot_graphviz(cf_graph)