cgraph: Casuality Graphs

In tzimiskes/causality: Examine The Causal Structure of Data

Description

Create or test for objects of type "causality.graph"

Usage

cgraph(nodes, edges, validate = TRUE)

is_valid_cgraph(graph)

is.cgraph(graph)

\method{summary}{causality.graph}(object, ...)

Arguments

nodes	A character array of node names
edges	A m x 3 character matrix. Each row is an edge in the form of (node1, node2, edgetype), with node1 and node2 being in nodes. Valid edge types are listed below
validate	logical value to determine whether or not to check to see if the cgraph is valid before returning it. Default is TRUE
graph	A graph to coerced or tested
object	A causality.graph
...	additional (unused) arguments to pass to summary

Details

A causality-graph consists of three things

• nodes: a character vector of the nodes of the in the causal graph

• adjacencies: a list of character vectors that contain the adjacencies of each node. This is calculated when a cgraph is created.

• edges: either a m x 3 character matrix or 3 x m character vector which represents the edges in a causal graph in the form (from, to, edge). For example, if we are dealing with a causal graph regarding drug use and cancer, The edge "Smoking –> Cancer" would be stored as ("Smoking", "Cancer", "–>") in the edge matrix.

The valid edges types for non latent variable model graphs (DAGs, PDAGs, Patterns) are:

• -->

• ---

And for latent variable models (PAGs, MAGs):

• o-o

• o->

• ++> (in Tetrad this is a green arrow)

• ~~> (in Tetrad this is a black arrow)

• <->

is_valid_cgraph checks to see if the input is a valid "causality.graph." Specifically, it checks that there are no duplicate nodes, self-loops, or multiple edges between pairs of nodes.

is.cgraph tests whether or not an object has the class causality.graph

summary provides basic summary statistics about graph, like average degree, max degree, number of directed/undirected eges etc.

Value

cgraph returns object of class "causality.graph", or an error if the graph is invalid.

is_valid_cgraph returns TRUE or FALSE depending on whether or not the input is valid.

is.cgraph returns TRUE or FALSE.

Author(s)

Alexander Rix

References

Spirtes et al. “Causation, Prediction, and Search.”, Mit Press, 2001, p. 109.

Spirtes P. Introduction to causal inference. Journal of Machine Learning Research. 2010;11(May):1643-62.

Pearl, Judea. Causality. Cambridge university press, 2009.

See Also

Other causality classes: dag, pattern coercing non causality graphs to causality.graphs : as.cgraph

Examples

nodes <- c("X1", "X2", "X3", "X4")
edges <- matrix(c("X1", "X2", "-->",
                  "X3", "X2", "-->",
                  "X4", "X1", "---",
                  "X4", "X3", "-->",
                  "X4", "X2", "-->"), ncol = 3, byrow = T)
graph <- cgraph(nodes, edges)

# cgraph defaults to validate = TRUE, but you can check validity by calling
is_valid_cgraph(graph)

# you can coerce graphs from package \code{bnlearn} to causality.graphs
## Not run: 
  library(bnlearn)
  sachs <- as.cgraph(mmhc(sachs.df))

## End(Not run)

tzimiskes/causality documentation built on Sept. 15, 2019, 8:41 p.m.