Evaluate a single fitted H3SDM model — h3sdm_eval

Computes performance metrics for a single fitted model, including standard yardstick metrics (ROC AUC, accuracy, sensitivity, specificity, F1-score, Kappa), as well as TSS (True Skill Statistic) and Boyce index. Designed as a helper function for evaluating models produced by h3sdm_fit_model or h3sdm_fit_models.

Usage

h3sdm_eval_metrics(
  fitted_model,
  presence_data = NULL,
  truth_col = "presence",
  pred_col = ".pred_1"
)

Arguments

fitted_model: A fitted model object, typically the result of h3sdm_fit_model().
presence_data: Optional. An sf object or tibble containing presence locations to compute the Boyce index.
truth_col: Character. Name of the column with true presence/absence values (default "presence").
pred_col: Character. Name of the column with predicted probabilities (default ".pred_1").

Value

A tibble with one row per metric, including:

.metric: Metric name (e.g., "roc_auc", "tss", "boyce").
.estimator: Estimator type (usually "binary").
mean: Metric value.
std_err: Standard error (NA for TSS and Boyce).
conf_low: Lower 95% confidence interval (NA for TSS and Boyce).
conf_high: Upper 95% confidence interval (NA for TSS and Boyce).