Dear Reader,
You will find in the associated text files ("novas-commandlines.R", "novas-routines.R",
and "novas-simulations.R") the materials needed for implementing the statistical methodology
detailed in the paper
FERRATY F., HALL P. An Algorithm for Nonlinear, Nonparametric Model Choice and Prediction
This method aims to select nonparametrically few variables among a large set of covariates
for predicting a scalar response.
The file "novas-routines.R" contains the whole set of R routines dealing with NOVAS
(NOnparametric Variable Selection) which includes several routines
("llreg", "novas", "novas.stepwise", and "predict.novas")
The file "novas-commandlines.R" contains all R commandlines allowing to implement the
nonparametric variable selection with different datasets:
==> petroleum prediction problem: given samples of gasoline, one predicts the octane number
by selecting nonparametrically few wavelengths involved in the corresponding near-infrared
spectrum;
==> food industry dataset: given a samples of meat, one predicts the fat content by
selecting nonparametrically few wavelengths involved in the corresponding near-infrared
spectrum;
==> genomics dataset: given a sample containing 3116 genes expressions and 9 clinical responses,
one predicts each clinical variable by selecting nonparametrically few genes expressions)
The file "novas-simulations.R" contains all R commandlines allowing to reproduce the results
given in TABLE 4 of the above-mentioned work.
How to use this material?
1) open an R session
2) follow the instructions given in the text file "novas-commandlines.R" or "novas-simulations.R"
(copy-paste the R commandlines into your R session)
Copyright: these materials may be downloaded freely for your own personal study. Use for any commercial purpose is forbidden.