sna (version 2.5)

cug.test: Univariate Conditional Uniform Graph Tests

Description

cug.test takes an input network and conducts a conditional uniform graph (CUG) test of the statistic in FUN, using the conditioning statistics in cmode. The resulting test object has custom print and plot methods.

Usage

cug.test(dat, FUN, mode = c("digraph", "graph"), cmode = c("size", 
    "edges", "dyad.census"), diag = FALSE, reps = 1000, 
    ignore.eval = TRUE, FUN.args = list())

Arguments

dat

one or more input graphs.

FUN

the function generating the test statistic; note that this must take a graph as its first argument, and return a single numerical value.

mode

graph if dat is an undirected graph, else digraph.

cmode

string indicating the type of conditioning to be applied.

diag

logical; should self-ties be treated as valid data?

reps

number of Monte Carlo replications to use.

ignore.eval

logical; should edge values be ignored? (Note: TRUE is usually more efficient.)

FUN.args

a list containing any additional arguments to FUN.

Value

An object of class cug.test.

Details

cug.test is an improved version of cugtest, for use only with univariate CUG hypotheses. Depending on cmode, conditioning on the realized size, edge count (or exact edge value distribution), or dyad census (or dyad value distribution) can be selected. Edges are treated as unvalued unless ignore.eval=FALSE; since the latter setting is less efficient for sparse graphs, it should be used only when necessary.

A brief summary of the theory and goals of conditional uniform graph testing can be found in the reference below. See also cugtest for a somewhat informal description.

References

Butts, Carter T. (2008). “Social Networks: A Methodological Introduction.” Asian Journal of Social Psychology, 11(1), 13--41.

See Also

cugtest

Examples

# NOT RUN {
#Draw a highly reciprocal network
g<-rguman(1,15,mut=0.25,asym=0.05,null=0.7)

#Test transitivity against size, density, and the dyad census
cug.test(g,gtrans,cmode="size")
cug.test(g,gtrans,cmode="edges")
cug.test(g,gtrans,cmode="dyad.census")
# }