pcalg (version 1.1-4)

udag2pdagSpecial: Extend a pcAlgo-object containing a skeleton to a PDAG using different methods if problems occur

Description

This function extends a pcAlgo-object containing a skeleton and corresponding conditional independence information to a Partially Directed Acyclic Graph (PDAG). The result is a pcAlgo-object as well.

Usage

udag2pdagSpecial(gInput,verbose,n.max=100)

Arguments

gInput

pcAlgo-object containing skeleton and cond. ind. information

verbose

0: No output; 1: Details

n.max

Maximum number of tries for reontienting doubly visited edges.

Value

• pcObj
  Oriented pc-Object
• evisit
  Matrix counting the number of orientation attemps per edge
• xtbl.orig
  Is original graph with v-structure extendable
• xtbl
  Is final graph with v-structure extendable
• amat0
  Adj.matrix of original graph with v-structures
• amat1
  Adj.matrix of graph with v-structures after reorienting edges from double edge visits
• status
  0: original try is extendable; 1: reorienting double edge visits helps; 2: orig. try is not extendable; reorienting double visits don't help; result is acyclic, has orig. v-structures, but perhaps additional v-structures
• counter
  Number of reorientation tries until success or max.tries

Details

The skeleton is extended to a PDAG using rules by Spirtes and Pearl (see References). If, after orienting the v-structures, the graph is not extendible to a DAG, the following mechanisms try to solve the problem: There might be edges, that were oriented twice while forming colliders. In this case, try all possible combinations of reorientations and check whether any reorientation is extendible. Take the first one that is extendible and make no more than n.max tries. If this fails, the original graph is extended arbitrarily to a DAG that fits on the skeleton. v-structures might have changed. The resulting DAG is then transformed to its CPDAG.

References

P. Spirtes, C. Glymour and R. Scheines (2000) Causation, Prediction, and Search, 2nd edition, The MIT Press.

J. Pearl (2000), Causality, Cambridge University Press.

See Also

pdag2dag, dag2cpdag, udag2pdag, udag2pdagRelaxed

Examples

## simulate data
set.seed(123)
p <- 10
myDAG <- randomDAG(p, prob = 0.2)
trueCPDAG <- dag2cpdag(myDAG)
n <- 1000
d.mat <- rmvDAG(n, myDAG, errDist = "normal")

## define independence test (partial correlations)
indepTest <- gaussCItest 
## define sufficient statistics
suffStat <- list(C = cor(d.mat), n = n)
## estimate skeleton
resU <- skeleton(suffStat, indepTest, p, 0.05)

## orient edges using three different methods
resD1 <- udag2pdagRelaxed(resU,verbose=0)
resD2 <- udag2pdagSpecial(resU,verbose=0,n.max=100)
resD3 <- udag2pdag(resU,verbose=0)