Package: nonlinvarsel
Title: Variable Selection For Nonlinear Models
Version: 0.0.1.9001
Authors@R: c(
    person("Robert", "McCulloch", email = "robert.e.mcculloch@gmail.com", role = c("aut", "cre")),
    person("Carlos", "Carvalho", email = "Carlos.Carvalho@mccombs.utexas.edu", role = "aut"),
    person("Richard", "Hahn", email = "prhahn@asu.edu", role = "aut"))
Description: We provide a general approach for variable selection 
    for nonlinear models such as BART (Bayesian Additive Regression Trees) 
    and deep neural nets.  We take the fit from the nonlinear model
    and find subsets of the variable such that we can approxiate the fit 
    using a nonlinear function of the subset.  In the case of BART, 
    where posterior draws of the nonlinear function are available, 
    we provide inference on the approximation error.  
    Our method emphasizes practical significance over statistical significance.
    We seek a subset of variables such that it is likely that we can 
    approximate the function using all the variables 
    with little error as a practical matter.
Depends: R (>= 2.14.0), foreach
Imports: rpart
Suggests: BART, knitr, MASS, nnet, rmarkdown
License: GPL (>= 2)
Encoding: UTF-8
LazyData: true
RoxygenNote: 6.1.1
NeedsCompilation: no
Packaged: 2020-04-11 21:28:30 UTC; rob
Author: Robert McCulloch [aut, cre],
  Carlos Carvalho [aut],
  Richard Hahn [aut]
Maintainer: Robert McCulloch <robert.e.mcculloch@gmail.com>
