itr: Individualized Treatment Rule Frameworks

In SkadiEye/ITRlearn: Individualized Treatment Rule Learning

Description

A collection of Individualized Treatment Rule frameworks.

Usage

itrDOWL(object, valid = NULL, regCtrl, wclsCtrl, ...)

itrOWL(object, valid = NULL, wclsCtrl, ...)

itrVT(object, valid = NULL, regCtrl, ...)

itrSimple(object, valid = NULL, regCtrl, ...)

EndLot(object, valid = NULL, ...)

LASSO(object, valid = NULL, ...)

OWL_LASSO(object, valid = NULL, ...)

VT_Single(object, valid = NULL, ...)

VT_Ensemble(object, valid = NULL, ...)

## S4 method for signature 'TrtDataObj'
itrDOWL(object, valid = NULL, regCtrl, wclsCtrl, ...)

## S4 method for signature 'TrtDataObj'
itrOWL(object, valid = NULL, wclsCtrl, ...)

## S4 method for signature 'TrtDataObj'
itrVT(object, valid = NULL, regCtrl, ...)

## S4 method for signature 'TrtDataObj'
itrSimple(object, valid = NULL, regCtrl, ...)

Arguments

object	A TrtDataObj object, used as the training set.
valid	A TrtDataObj object, used as the validation set. The default is NULL.
regCtrl	A list of arguments that specifies a regression method, handled by regCtrlPanel.
wclsCtrl	A list of arguments that specifies a weighted classification method, handled by wclsCtrlPanel.
...	Other parameters passed to itr functions.

Details

These itr frameworks all works for the personalized treatment problems with two treatmnet labels. Different frameworks would call either regression methods or classification methods, or both. Methods of the same type could be applied in the same framework.

Value

Returns a ITRObj object.

Methods (by generic)

itrDOWL: Doubly Weighted Outcome Learning Framework. This procedure first calls a regression model to estimate E(Y|X), the expected treatment effect averaged on all treatments. Then it calls a weighted classification to estimate the ITR.

itrOWL: Outcome Weighted Learning Framework. This procedure directly calls a weighted classification to estimate the ITR by outcome weighted learning.

itrVT: Virture Twins Framework. This procedure calls the same method to fit two regression models for patients with either treatment. The ITR is determined by comparing the predictions with both models.

itrSimple: Simple Regression Framework. This procedure fits one regression model with A*Y ~ X, where A from 1, -1 is the treatment label, Y is the continuous outcome and X is the predictors.

References

Outcome Weighted Learning
Virtual Twins
A simple Regression

See Also

TrtDataObj-class
ITRObj-class
regCtrlPanel
wclsCtrlPanel
reg
wcls
cvGrid
ensemble
cvKfold
workflow

SkadiEye/ITRlearn documentation built on May 24, 2019, 1:31 a.m.