pcalg (version 2.7-12)

compareGraphs: Compare two graphs in terms of TPR, FPR and TDR

Description

Compares the true undirected graph with an estimated undirected graph in terms of True Positive Rate (TPR), False Positive Rate (FPR) and True Discovery Rate (TDR).

Usage

compareGraphs(gl, gt)

Value

A named numeric vector with three numbers:

tpr

True Positive Rate: Number of correctly found edges (in estimated graph) divided by number of true edges (in true graph)

fpr

False Positive Rate: Number of incorrectly found edges divided by number of true gaps (in true graph)

tdr

True Discovery Rate: Number of correctly found edges divided by number of found edges (both in estimated graph)

Arguments

gl

Estimated graph (graph object)

gt

True graph (graph object)

Author

Markus Kalisch () and Martin Maechler

Details

If the input graph is directed, the directions are omitted. Special cases:

• If the true graph contains no edges, the tpr is defined to be zero.

• Similarly, if the true graph contains no gaps, the fpr is defined to be one.

• If there are no edges in the true graph and there are none in the estimated graph, tdr is one. If there are none in the true graph but there are some in the estimated graph, tdr is zero.

See Also

randomDAG for generating a random DAG.

Examples

## generate a graph with 4 nodes
V <- LETTERS[1:4]
edL2 <- vector("list", length=4)
names(edL2) <- V
edL2[[1]] <- list(edges= 2)
edL2[[2]] <- list(edges= c(1,3,4))
edL2[[3]] <- list(edges= c(2,4))
edL2[[4]] <- list(edges= c(2,3))
gt <- new("graphNEL", nodes=V, edgeL=edL2, edgemode="undirected")

## change graph
gl <- graph::addEdge("A","C", gt,1)

## compare the two graphs
if (require(Rgraphviz)) {
par(mfrow=c(2,1))
plot(gt) ; title("True graph")
plot(gl) ; title("Estimated graph")
(cg <- compareGraphs(gl,gt))
# \dontshow{
stopifnot(cg == c(1, 1/2, 4/5))
# }
}