corGraph: Computing the correlation graph

In pcalg: Methods for Graphical Models and Causal Inference

Computing the correlation graph

Description

Computes the correlation graph. This is the graph in which an edge is drawn between node i and node j, if the null hypothesis “Correlation between X_i and X_j is zero” can be rejected at the given significance level \alpha (alpha).

Usage

corGraph(dm, alpha=0.05, Cmethod="pearson")

Arguments

dm	numeric matrix with rows as samples and columns as variables.
alpha	significance level for correlation test (numeric)
Cmethod	a character string indicating which correlation coefficient is to be used for the test. One of "pearson", "kendall", or "spearman", can be abbreviated.

Value

Undirected correlation graph, a graph object (package graph); getGraph for the “fitted” graph.

Author(s)

Markus Kalisch ([e-mail]) and Martin Maechler

Examples

## create correlated samples
x1 <- rnorm(100)
x2 <- rnorm(100)
mat <- cbind(x1,x2, x3 = x1+x2)

if (require(Rgraphviz)) {
## ``analyze the data''
(g <- corGraph(mat)) # a 'graphNEL' graph, undirected
plot(g) # ==> (1) and (2) are each linked to (3)

## use different significance level and different method
(g2 <- corGraph(mat, alpha=0.01, Cmethod="kendall"))
plot(g2) ## same edges as 'g'
}

pcalg documentation built on Feb. 6, 2024, 3 p.m.