Fit and Analyze Smooth Supersaturated Models

Creates an S4 class "SSM" and defines functions for fitting smooth supersaturated models, a polynomial model with spline-like behaviour. Functions are defined for the computation of Sobol indices for sensitivity analysis and plotting the main effects using FANOVA methods. It also implements the estimation of the SSM metamodel error using a GP model with a variety of defined correlation functions.

The SSM package provides functions to fit, plot and predict using smooth supersaturated models. It defines an S4 class called "SSM", and methods for plotting and predicting them. The fitting function is highly customizable and provides optional sensitivity analysis and the provision to estimate metamodel error using a Gaussian process.

The following code will fit a smooth supersaturated model to a 20 point design in four factors. Note the design should be held in a matrix, not a data.frame, and all entries must be numeric. The options SA, GP and validation turn on automated sensitivity analysis, Gaussian process metamodel error estimation and Leave-One-Out cross-validation respectively. The plot method plots the main effects of the model while the predict method gives the model prediction at a point and also a 95% credible interval if a metamodel error GP has been fit.

X <- matrix(runif(80, -1, 1), ncol = 4)
Y <- apply(apply(X, 1, "^", 1:4), 2, sum)
s <- fit.ssm(X, Y, SA = TRUE, GP = TRUE, validation = TRUE)
plot(s, yrange="yrange")
predict(s, rep(0.5, 4))

To install the most up-to-date SSM package through GitHub use devtools::install_github("peterrobertcurtis/SSM").

More details on how to use the SSM can be found in the vignette and help pages.


Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


1.0.1 by Peter Curtis, 2 years ago

Report a bug at

Browse source code at

Authors: Peter Curtis [aut, cre]

Documentation:   PDF Manual  

GPL-3 license

Imports methods

Suggests knitr, rmarkdown

See at CRAN