SAVE: Sliced Average Variance Estimator

In abnormally-distributed/cvreg: Cross Validation and Robust Estimation Utilities

Description

A problem that SIR, KIR, and PIR face is that they cannot recover a vector in the central subspace if the regression function is symmetric about 0. To improve the situation, methods based on second-order conditional moments have been developed to improve the difficulty. The sliced average variance estimator (SAVE) estimates the conditional expectation of the transposed crossproduct of X given Y, E(XX'| Y). Like SIR, SAVE also splits Y into bins and estimates the inverse regression as a piecewise function (Cook & Weisburg, 1991). Since SAVE is sensitive to directions that first order moment methods such as SIR cannot detect, the central subspace estimated by SAVE compared to them will tend to be larger to or equal in size.

Usage

SAVE(
  formula,
  data,
  rank = "all",
  slices = 4,
  ytype = c("numeric", "categorical")
)

Arguments

formula	a model formula
data	a data frame
rank	the desired number of sufficient predictors to return. the default is "all".
slices	the number of slices into which the response variable should be split. defaults to 4. for categorical response variables the maximum allowed is the number of response levels minus one. if set above this, it is silently adjusted.
ytype	either numeric or categorical

Value

an sdr object

References

Cook, R.D., Weisberg, S. (1991) Sliced Inverse Regression for Dimension Reduction: Comment. Journal of the American Statistical Association, 86(414):328-332

abnormally-distributed/cvreg documentation built on May 3, 2020, 3:45 p.m.