Compare estimated graph to ground truth

Shows side-by-side heatmaps of the true adjacency matrix, PIP, and thresholded estimate. Returns confusion metrics invisibly.

Usage

plot_recovery(fit, true_adj, pip_threshold = 0.5, groups = NULL)

Arguments

fit

A multiggm_fit object returned by multiggm_mcmc.

true_adj

A list of K true adjacency matrices (0/1), or a single matrix (recycled for all groups).

pip_threshold

Numeric; threshold for edge selection. Default 0.5.

groups

Integer vector; which groups to plot. Default: all groups.

Value

A named list of confusion metric vectors per group (invisible). Each element is a named numeric vector as returned by confusion_at_threshold, with components TP, FP, TN, FN, TPR, FPR.

See also

confusion_at_threshold(), roc_auc(), plot_roc()

Examples

sim <- simulate_multiggm(K = 2, p = 8, n = 80, seed = 1)
fit <- multiggm_mcmc(data_list = sim$data_list, burnin = 200, nsave = 100)
cm <- plot_recovery(fit, sim$adj_list)

cm$Group_1  # confusion metrics for group 1
#> $TP
#> [1] 8
#> 
#> $FP
#> [1] 0
#> 
#> $TN
#> [1] 15
#> 
#> $FN
#> [1] 5
#> 
#> $TPR
#> [1] 0.6153846
#> 
#> $FPR
#> [1] 0
#> 
#> $precision
#> [1] 1
#> 
#> $F1
#> [1] 0.7619048
#>