Package: GLAMMGoF 1.1.2

GLAMMGoF: Resampling-Based Predictive Validation for Generalized Linear and Generalized Additive Models

Provides resampling-based predictive validation for generalized linear and generalized additive models (with or without random effects) fitted using packages such as glmmTMB, mgcv, stats, lme4, and MASS. Predictive performance is assessed using either repeated random holdout (Monte Carlo cross-validation) or bootstrap resampling with out-of-bag evaluation. In each replicate, models are refit to a training dataset and evaluated on separate testing data, generating sampling distributions of in-sample and out-of-sample performance statistics. For continuous or integer response models, supported metrics include relative root mean squared error (RRMSE), relative mean absolute error (RMAE), relative median absolute error (RMedAE), and relative bias (RBIAS). For binary response models, supported metrics include AUC, Brier score, and log loss. All predictive metrics are based on population-level predictions, meaning random effects are excluded when present. Optional residual diagnostics can also be performed using the DHARMa package.

Authors:Colin Shea [aut, cre]