R bnlearn Grow-Shrink structure learning returns undirected graph

Question

The Nagarajan et al. book (Bayesian Networks in R, O'Reilly 2013, p. 35) says that when I take the marks dataset of the R bnlearn package and ask to learn structure using the grow-shrink implementation by writing

library(bnlearn)
data(marks)
bn.gs = gs(marks)

then I should obtain a directed graph:

model: [STAT][ANL|STAT][ALG|ANL:STAT][VECT|ALG] [MECH|VECT:ALG]
nodes: 5
arcs: 6
undirected arcs: 0
directed arcs: 6

Instead of that I get an undirected graph:

model: [undirected graph]
nodes: 5 
arcs: 6 
undirected arcs: 6 
directed arcs: 0 

Even when I add the option undirected=FALSE to the gs method, I still get the same result. What am I doing wrong? Or is there a bug in the R implementation? I have tried it both on my Mac and on a Debian machine, the result is the same...

Answer 1

The answer is provided by the authors on their Web site (which I just discovered): http://www.bnlearn.com/book-useR/

“page 35: bnlearn 3.2 and later versions are more picky about setting arc directions; as a result bn.gs is an undirected graph and must be extended into a DAG with cextend() to conclude the example.”

In other words, replace the last line of code by

bn.gs = cextend(gs(marks))

It is a pity that Springer does not provide a standard service for errata, as does, for example, O'Reilly. The authors of this book had to create their own Web site, and it took my some time to find it...