Well influence and network thinning

Leave-one-well influence compares fixed-geometry reduced fits with a full reference and extracts the held-out residual. Influence can arise from location, network geometry, local gradient, unique hydrogeologic information, or an anomalous value; it does not classify a well as erroneous. Network thinning separates held-out predictive error from descriptive difference against the full-network surface and records duplicate retained sets.

influence <- ps_well_influence(points, "TPS", grid_res=100)
thin <- ps_network_thinning(points, retain=c(.75,.5), repeats=20, seed=5)