rie.d: Relative Information Error using data sets as input

In ShuoStat/RIE: Estimate the relative information error of selected models

Relative Information Error using data sets as input

Description

Calculate the information error and relative information error using the data as the input. The data will be transformed to discretized data, from where the probability is calculated.

Usage

rie.d(
  X,
  rvars,
  svars,
  contivars = NULL,
  method = "emp",
  disc = "equalfreq",
  nbins = nrow(X)^(1/3)
)

Arguments

X	matrix or data.frame containing all variables
rvars	relevant variables
svars	selected variables
contivars	which variables are continuous? By default, only continuous variables will be discretized. If NULL, all variables are continuous.
method	the name of the entropy estimator. We use the entropy computation of the infotheo package, where four methods are includes: "emp", "mm", "shrink", "sg". For more details, see entropy() in "infotheo" package.
disc	the strategies to discretize data. Three options are provided: "equalfreq", "equalwidth", and "globalequalwidth". See discretize() in "infotheo" package.
nbins	iteger specifying the number of bins to be used for the discretization

Author(s)

Shuo Wang

Examples

m <- GenMatrix(3, rlist = list(c(1, 2, 0.5), c(1, 3, 0.3), c(2, 3, 0.7)))
D <- GenData(100, rmatrix = m)
rvars = c("x1", "x2")
svars = c("x2", "x3")
rie.d(D, rvars, svars)

ShuoStat/RIE documentation built on March 19, 2022, 3:30 a.m.