Per-element pair concordance score — getPairConcordance • poem

Per-element pair concordance between a clustering and a ground truth. Note that by default, negative pairs (i.e. that are split in both the predicted and true groupings) are not counted. To count it (as in the standard Rand Index), use useNegatives=TRUE.

Usage

getPairConcordance(
  true,
  pred,
  usePairs = TRUE,
  useNegatives = FALSE,
  adjust = FALSE
)

Arguments

true: A vector of true class labels
pred: A vector of predicted clusters
usePairs: Logical; whether to compute over pairs instead of elements Recommended and TRUE by default.
useNegatives: Logical; whether to include the consistency of negative pairs in the score (default FALSE).
adjust: Logical; whether to adjust for chance. Only implemented for useNegatives=FALSE (doesn't make sense on a element-level otherwise).

Value

A vector of concordance scores