Compute confusion matrix metrics at a threshold

Given a score matrix (e.g., edge inclusion probabilities) and a binary truth matrix, computes TP, FP, TN, FN, TPR, FPR, precision, and F1 at a given threshold. Only the upper triangle is used by default.

Usage

confusion_at_threshold(
  score_mat,
  truth_mat,
  threshold = 0.5,
  upper_only = TRUE
)

Arguments

score_mat

Numeric matrix of scores (e.g., probabilities in [0,1]).

truth_mat

Binary matrix of true edges (1 = edge, 0 = no edge).

threshold

Numeric threshold; scores >= threshold are predicted as edges. Default 0.5.

upper_only

Logical; if TRUE (default), only use upper triangle.

Value

A named list with elements: TP, FP, TN, FN, TPR (sensitivity), FPR (1-specificity), precision, F1.

Examples

# Simulated score and truth matrices
truth <- matrix(c(0,1,0, 1,0,1, 0,1,0), 3, 3)
scores <- matrix(c(0,0.9,0.1, 0.9,0,0.8, 0.1,0.8,0), 3, 3)
confusion_at_threshold(scores, truth, threshold = 0.5)
#> $TP
#> [1] 2
#> 
#> $FP
#> [1] 0
#> 
#> $TN
#> [1] 1
#> 
#> $FN
#> [1] 0
#> 
#> $TPR
#> [1] 1
#> 
#> $FPR
#> [1] 0
#> 
#> $precision
#> [1] 1
#> 
#> $F1
#> [1] 1
#>