Posterior probability of differential partial correlation (K=2)

For two-group models, computes the posterior probability that the partial correlation between each pair of variables differs between groups by more than a threshold delta.

Usage

diff_prob_pcor(pcor, delta = 0)

Arguments

pcor

A numeric array of dimension [p, p, 2, nsave], as returned by posterior_pcor with K = 2.

delta

Numeric; practical difference threshold. Default 0 (any difference). Set to e.g. 0.1 to detect meaningful differences only.

Value

A symmetric numeric matrix [p, p] where entry [i,j] is the posterior probability that \(|\rho_{ij}^{(1)} - \rho_{ij}^{(2)}| > \delta\).

See also

posterior_pcor(), pip_diff_edge()

Examples

sim <- simulate_multiggm(K = 2, p = 8, n = 80, seed = 1)
fit <- multiggm_mcmc(data_list = sim$data_list, burnin = 200, nsave = 100)
pcor_draws <- posterior_pcor(fit)
diff_prob <- diff_prob_pcor(pcor_draws, delta = 0.1)
# Edges with high probability of differential partial correlation
sum(diff_prob[upper.tri(diff_prob)] > 0.5)
#> [1] 10