Routines for generalized additive modelling under shape constraints on the component functions of the linear predictor (Pya and Wood, 2015) <doi:10.1007/s11222-013-9448-7>. Models can contain multiple shape constrained (univariate and/or bivariate) and unconstrained terms. The routines of gam() in package 'mgcv' are used for setting up the model matrix, printing and plotting the results. Penalized likelihood maximization based on Newton-Raphson method is used to fit a model with multiple smoothing parameter selection by GCV or UBRE/AIC.
This version of scam can be found in the following snapshots: