K-Fold Cross Validation for FishSET Models

Performs k-fold cross-validation on a fitted FishSET model to evaluate out-of-sample predictive performance. Bypasses formula parsing by directly subsetting the design matrix.

Usage

fishset_cv(
  project,
  base_model_name,
  k = 5,
  seed = 42,
  distribution = NULL,
  ...
)

Arguments

project: Character string. Name of the project.
base_model_name: Character string. The name of the original model design to cross-validate.
k: Integer. The number of folds to create. Default is 5.
seed: Integer. Random seed for reproducible fold generation. Default is 42.
distribution: Character string. Distribution for the continuous catch component in EPMs.
...: Additional control arguments passed to fishset_fit().

Value

A list containing the average out-of-sample accuracy, log-likelihood, PAPE, AIC, fold details, and estimated coefficients across folds.

Examples

if (FALSE) { # \dontrun{
# Standard logit
cv_results <- fishset_cv(
  project = "MyProject",
  base_model_name = "clogit_design"
)

# EPM - normal catch function
cv_epm_results <- fishset_cv(
  project = "MyProject",
  base_model_name = "epm_design",
  distribution = "normal"
)
} # }