Summarizes each chain and the pooled posterior. The pooled edge counts use posterior inclusion probabilities averaged over all chains. The largest difference between two chains' PIPs (or, for mGHS, posterior mean partial correlations) is a quick convergence check: chains that agree give small values.
Usage
# S3 method for class 'multiggm_fit_list'
summary(object, pip_threshold = 0.5, ...)Arguments
- object
A
multiggm_fit_listreturned bymultiggm_mcmcwithnchains > 1.- pip_threshold
Numeric; PIP threshold for the edge counts. Default 0.5.
- ...
Ignored.
Value
An object of class "summary.multiggm_fit_list": a list with
chains (the per-chain summaries), edge_counts (pooled, per
group, or NULL for mGHS), max_chain_diff (largest absolute
between-chain difference, per group), nchains, K and
pip_threshold.
Examples
sim <- simulate_multiggm(K = 2, p = 6, n = 60, seed = 1)
fit <- multiggm_mcmc(data_list = sim$data_list, burnin = 100, nsave = 50,
nchains = 2, seed = 1)
summary(fit)
#> multiGGM MCMC Summary: 2 chains, K = 2
#> ==========================================
#> Pooled selected edges (PIP >= 0.5 ):
#> Group 1: 7 edges
#> Group 2: 7 edges
#>
#> Largest difference in PIP between chains:
#> Group 1: 0.480
#> Group 2: 0.520
#>
#> --- Chain 1 ---
#> multiGGM MCMC Summary
#> =====================
#> Groups (K): 2 | Nodes (p): 6 | Posterior draws: 50
#>
#> Acceptance rates:
#> gamma (edge toggle): 6.0%
#> theta (within-model): 36.3%
#> nu (edge log-odds): 58.5%
#>
#> Selected edges (PIP >= 0.5 ):
#> Group 1 : 8 edges
#> Group 2 : 8 edges
#>
#> Graph similarity (theta):
#> theta[1,2]: mean = 1.022, P(nonzero) = 98.0%
#>
#>
#> --- Chain 2 ---
#> multiGGM MCMC Summary
#> =====================
#> Groups (K): 2 | Nodes (p): 6 | Posterior draws: 50
#>
#> Acceptance rates:
#> gamma (edge toggle): 2.0%
#> theta (within-model): 39.6%
#> nu (edge log-odds): 59.4%
#>
#> Selected edges (PIP >= 0.5 ):
#> Group 1 : 7 edges
#> Group 2 : 7 edges
#>
#> Graph similarity (theta):
#> theta[1,2]: mean = 0.764, P(nonzero) = 98.0%
#>