This function normalized the covariates in the data to lie in
the unit interval based on either the empirical or known ranges of the
covariates. It is useful to perform this step when defining the instrumental
functions later on. This function is used in G.box.R, G.spline.R
and by extension in G.cd.R.
normalize.covariates(
data = NULL,
x = NULL,
cov.ranges = NULL,
idxs.c = "all",
norm.cov.out = NULL,
...
)(optional) Data set to be used to construct the normalizing
transformation. Default is data = NULL.
(optional) Vector of covariates to be normalized alongside the data.
Default is x = NULL.
(optional) Matrix that specifies the range of each of the
covariates in the data set. Each column corresponds to a covariate. The first
row contains the lower bound, the second row contains the upper bound.
If not supplied, the data will be normalized based on the minimum and maximum
detected values. If supplied, the non data-dependent transformation function
listed in the appendix of Andrews, Shi 2013 will be used. Default is
cov.ranges = NULL.
(optional) Vector of indices of covariates that are continuous.
Note that that indices are relative to the covariate vector, not the full
data set. Default value is idxs.c = "all", which indicates that all
elements should be regarded as continuous. If idxs.c = NULL, all
elements are regarded as discrete.
(optional) The output of a previous call to this function.
Can be used to speed up computation. If both data and
norm.cov.out are supplied to the function, this method will throw an
error. Default is norm.cov.out = NULL.
Allows easier interchangeability between covariate normalization
functions. All arguments specified under ... will be ignored.
Andrews, D.W.K. and Shi, X. (2013). Inference based on confitional moment inequalities. Econometrica. 81(2):609-666.