Skip to contents

Posterior inclusion probabilities (PIP) per group

pip_edges.Rd

Computes the posterior inclusion probability for each edge in each group. The PIP is the fraction of saved MCMC iterations in which each edge was included in the graph.

Usage

pip_edges(fit, chain = 1L, platform = NULL)

Arguments

fit

A multiggm_fit object (single chain), multiggm_fit_list object, or multiggm_platform_fit object.

chain

Integer; which chain to use if fit is a multiggm_fit_list. Default 1.

platform

Integer; which platform to extract (required for multiggm_platform_fit objects). Ignored for single-platform fits.

Value

A numeric array of dimension [p, p, K]. Entry pip[i, j, k] is the posterior probability that edge (i,j) is present in the graph of group k. Values range from 0 to 1. The matrix is symmetric for each group (pip[i,j,k] = pip[j,i,k]). Diagonal entries are 0.

See also

pip_diff_edge(), confusion_at_threshold(), roc_auc()

Examples

sim <- simulate_multiggm(K = 2, p = 8, n = 80, seed = 1)
fit <- multiggm_mcmc(data_list = sim$data_list, burnin = 200, nsave = 100)
pip <- pip_edges(fit)
# Edges with PIP >= 0.5 in group 1
sum(pip[,,1][upper.tri(pip[,,1])] >= 0.5)
#> [1] 8