christithomp/scamplot: Predictive Plots for Shape-Constrained Additive Models

This package is designed to create predictive plots for logistic shape-constrained additive models (SCAM) developed in the R package scam. Developing a predictive plot requires changing all variables (factors included) into numeric variables and refitting the SCAM with these new variables. Once this step is complete, new data frames must be developed for each variable modeled as a spline. All covariates and spline terms except for the targeted one in the new data frames are set to their mean values. The targeted spline terms are sequenced from their minimum to maximum values. Once the data frames are developed, the predict() function can be used to model either the logits or the probabilities, and the respective plots can be created from here. This package is useful because adding in a monotonic or concavity/convexity constraint, you must first be able to visualize your spline terms to see which shape constraint should be applied.

install.packages("remotes")
remotes::install_github("christithomp/scamplot")