sdr: sdr: methods for sufficient dimension reduction

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

Description

This defines the methods for the class "sdr", which is utilized for sufficient dimension reduction. Sufficient dimension reduction is an approach for reducing the dimensionality of a set of covariates based on the statistical principle of sufficiency. A sufficient statistic is a number that captures all of the information about a data set. Likewise, a set of sufficient predictors capture all of the information about an outcome Y contained in the set of covariates X. Sufficient predictors can be thought of as similar to principal components.

Several methods are offered in this package. Many codes here are based upon Li (2018), with modifications and improvements for more efficient calculation, and the addition of new features. Also included are methods for partial sufficient dimension reduction, which estimates sufficient dimension reduction subspaces integrated over a 'nuisance variable', typically a categorical predictor. In addition, envelope models, regularized SIR and SAVE, structured linear regression (which allows predictors to be grouped), and principle fitted components are also offered.

Though not of the sdr class, the estimators in the functions gpcr and gpls are geared towards the same goal as the sufficient dimension reduction approach.

Usage

sdr(x, ...)

Arguments

x	argument
...	other arguments

Details

sdr: methods for sufficient dimension reduction

References

Li, B. (2018). Sufficient dimension reduction: methods and applications with R. Boca Raton: CRC Press, Taylor & Francis Group.

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