threeboost-package: Thresholded boosting based on estimating equations

In threeboost: Thresholded variable selection and prediction based on estimating equations

Description

The threeboost package implements a the EEBoost algorithm described in Wolfson (2011, JASA). EEBoost is a general-purpose method for variable selection which can be applied whenever inference would be based on an estimating equation. Thresholded EEBoost (function threeboost) is a generalization of EEBoost which allows multiple variables to enter the model at each boosting step. EEBoost (function eeboost) is a special case of thresholded boosting with the threshold set to 1.

The package currently provides a "pre-packaged" function, geeboost, which carries out variable selection for correlated outcome data based on the Generalized Estimating Equations. However, the threeboost (and eeboost) functions can also accommodate user-provided estimating functions.

threeboost documentation built on May 2, 2019, 2:37 a.m.