methods-coef: Extract Model Coefficients

In DynTxRegime: Methods for Estimating Dynamic Treatment Regimes

Description

Extract model coefficients from regression analyses.

Usage

 ## S4 method for signature 'DynTxRegime'
coef(object, ...)  

Arguments

object	an object of class DynTxRegime.
...	passed through to coef() method of modelObj.

Details

There is a pattern to the format for the returned values of all standard regression tools that have been extended to the DynTxRegime class. A list is always returned. The length and structure of the list are determined by the statistical method used to estimate the optimal dynamic treatment regime and by the combination of the input arguments.

For IQ-Learning, only outcome regression models are used. These models are always communicated using objects of class modelObj. The pattern for the returned object follows:

if( iter == 0 && class(moMain) == "modelObj" && class(moCont) == "modelObj")

names(coef()) = "Combined"

if( iter > 0 && class(moMain) == "modelObj" && class(moCont) == "modelObj")

names(coef()) = "MainEffect", "Contrast"

if( iter == 0 && class(moMain) == "modelObj" && class(moCont) == "NULL")

names(coef()) = "moMain"

if( iter == 0 && class(moMain) == "NULL" && class(moCont) == "modelObj")

names(coef()) = "moCont"

For Q-Learning, only outcome regression models are used. If regression models are defined using objects of class modelObj, the returned list will follow the pattern described for IQ-Learning. If regression models are defined using objects of class modelObjSubset, a list is returned. The kth element of that list corresponds to the kth subset model and is named accordingly. Each subset model follows the pattern for modelObj described above.

For optimalSeq, a list is returned, the elements of which are named "regimes," "propen," and "outcome." Element "regimes" will contain the parameter estimates for the treatment regimes. Element "propen" will contain the parameter estimates for each of the propensity for treatment models. And, "outcome" will contain the parameter estimates for each of the outcome regression models.

For optimalClass, a list is returned, the elements of which are named "propen" and "outcome." Element "propen" will contain the parameter estimates for the propensity for treatment model. And, "outcome" will contain the parameter estimates for the outcome regression models. The coef method is not appropriate for classification regression and is thus excluded from the returned list.

Value

Returns a list. See the details section for more information.

Author(s)

Shannon T. Holloway <[e-mail]>

DynTxRegime documentation built on May 2, 2019, 5:21 p.m.