Changes in version 0.1.1 (2026-11-17) First release. - gam_influence() deletes each site (cluster) of a gam() fit in turn and refits the model twice, with the estimated smoothing parameters (and negative binomial theta) held and re-estimated. It reports the fixed, added and total changes in the other sites' fitted values, and a table of hyperparameter changes. - The re-estimated refit is searched from two starting points and the better REML fit is kept; status flags sites with two clearly different optima. - plot() shows each site's fixed change against its added change. - A scope guard rejects models outside the supported class, and a self-check confirms that the refits reproduce the original fit. - Dataset coral_trout: 467 surveys at 71 inshore reef sites of the Great Barrier Reef, derived from Australian Institute of Marine Science data under CC BY 3.0 Australia (see LICENSE.note). - Vignette: "Which sites change a GAM, and why? A coral trout example".