netrankr (version 1.2.4)

exact_rank_prob: Probabilistic centrality rankings

Description

Performs a complete and exact rank analysis of a given partial ranking. This includes rank probabilities, relative rank probabilities and expected ranks.

Usage

exact_rank_prob(P, only.results = TRUE, verbose = FALSE, force = FALSE)

Value

lin.ext

Number of possible rankings that extend P.

mse

Array giving the equivalence classes of P.

rank.prob

Matrix containing rank probabilities: rank.prob[u,k] is the probability that u has rank k.

relative.rank

Matrix containing relative rank probabilities: relative.rank[u,v] is the probability that u is ranked lower than v.

expected.rank

Expected ranks of nodes in any centrality ranking.

rank.spread

Standard deviation of the ranking probabilities.

topo.order

Random ranking used to build the lattice of ideals (if only.results = FALSE).

tree

Adjacency list (incoming) of the tree of ideals (if only.results = FALSE).

lattice

Adjacency list (incoming) of the lattice of ideals (if only.results = FALSE).

ideals

List of order ideals (if only.results = FALSE).

In all cases, higher numerical ranks imply a higher position in the ranking. That is, the lowest ranked node has rank 1.

Arguments

P

A partial ranking as matrix object calculated with neighborhood_inclusion or positional_dominance.

only.results

Logical. return only results (default) or additionally the ideal tree and lattice if FALSE.

verbose

Logical. should diagnostics be printed. Defaults to FALSE.

force

Logical. If FALSE (default), stops the analysis if the partial ranking has more than 40 elements and less than 0.4 comparable pairs. Only change if you know what you are doing.

Author

David Schoch, Julian Müller

Details

The function derives rank probabilities from a given partial ranking (for instance returned by neighborhood_inclusion or positional_dominance). This includes the calculation of expected ranks, (relative) rank probabilities and the number of possible rankings. Note that the set of rankings grows exponentially in the number of elements and the exact calculation becomes infeasible quite quickly and approximations need to be used. See vignette("benchmarks") for guidelines and approx_rank_relative, approx_rank_expected, and mcmc_rank_prob for approximative methods.

References

De Loof, K. 2009. Efficient computation of rank probabilities in posets. Phd thesis, Ghent University.

De Loof, K., De Meyer, H. and De Baets, B., 2006. Exploiting the lattice of ideals representation of a poset. Fundamenta Informaticae, 71(2,3):309-321.

See Also

approx_rank_relative, approx_rank_expected, mcmc_rank_prob

Examples

P <- matrix(c(0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, rep(0, 10)), 5, 5, byrow = TRUE)
P
res <- exact_rank_prob(P)

# a warning is displayed if only one ranking is possible
tg <- threshold_graph(20, 0.2)
P <- neighborhood_inclusion(tg)
res <- exact_rank_prob(P)