| Title: | Methods for Estimating Optimal Dynamic Treatment Regimes |
|---|---|
| Description: | Methods to estimate dynamic treatment regimes using Interactive Q-Learning, Q-Learning, weighted learning, and value-search methods based on Augmented Inverse Probability Weighted Estimators and Inverse Probability Weighted Estimators. Dynamic Treatment Regimes: Statistical Methods for Precision Medicine, Tsiatis, A. A., Davidian, M. D., Holloway, S. T., and Laber, E. B., Chapman & Hall/CRC Press, 2020, ISBN:978-1-4987-6977-8. |
| Authors: | S. T. Holloway, E. B. Laber, K. A. Linn, B. Zhang, M. Davidian, and A. A. Tsiatis |
| Maintainer: | Shannon T. Holloway <[email protected]> |
| License: | GPL-2 |
| Version: | 4.15 |
| Built: | 2025-01-03 05:43:59 UTC |
| Source: | https://github.com/cran/DynTxRegime |
A dataset generated to mimic data from a two-stage randomized clinical trial that studied the effect of meal replacement shakes on adolescent obesity. The dataset contains the following covariates collected at the start of the first stage: "gender," "race," "parentBMI," and "baselineBMI." At the second-stage, "month4BMI" was collected. Variables "A1" and "A2" are the randomized treatments at stages one and two, and "month12BMI" is the primary outcome collected at the end of stage two.
A matrix with rows corresponding to patients.
Generated by Kristin A. Linn in R
Function performs a single step of the bowl method. Multiple decision points can be analyzed by repeated calls, as is done for qLearn() and optimalClass().
bowl( ..., moPropen, data, reward, txName, regime, response, BOWLObj = NULL, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, fSet = NULL, surrogate = "hinge", verbose = 2L )bowl( ..., moPropen, data, reward, txName, regime, response, BOWLObj = NULL, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, fSet = NULL, surrogate = "hinge", verbose = 2L )
... |
Used primarily to require named input. However, inputs for the optimization methods can be sent through the ellipsis. If surrogate is hinge, the optimization method is dfoptim::hjk(). For all other surrogates, stats::optim() is used. |
moPropen |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the propensity for tx. See ?moPropen for details. |
data |
A data frame of the covariates and tx histories. |
reward |
The response vector. |
txName |
A character object. The column header of data that corresponds to the tx covariate |
regime |
A formula object or a list of formula objects. The covariates to be included in the decision function/kernel. If a list is provided, this specifies that there is an underlying subset structure – fSet must then be defined. For subsets, the name of each element of the list must correspond to the name of a subset. If a regime is to be estimated using multiple subsets combined, each subset must be included in the name and separated by a comma (no spaces). |
response |
A numeric vector. The same as reward above. Allows for naming convention followed in most DynTxRegime methods. |
BOWLObj |
NULL or |
lambdas |
A numeric object or a numeric vector object giving the penalty tuning parameter(s). If more than 1 is provided, the set of tuning parameter values to be considered in the cross-validation algorithm (note that cvFolds must be positive in this case). |
cvFolds |
If cross-validation is to be used to select the tuning parameters and/or kernel parameters, the number of folds. |
kernel |
A character object. Must be one of {'linear', 'poly', 'radial'} |
kparam |
A numeric object. |
fSet |
A function or NULL defining subset structure. See ?fSet for details. |
surrogate |
The surrogate 0-1 loss function. Must be one of {'logit', 'exp', 'hinge', 'sqhinge', 'huber'}. |
verbose |
An integer or logical. If 0, no screen prints are generated. If 1, screen prints are generated with the exception of optimization results obtained in iterative algorithm. If 2, all screen prints are generated. |
a BOWL-class object
Yingqi Zhao, Donglin Zeng, Eric B. Laber, Michael R. Kosorok (2015) New statistical learning methods for estimating optimal dynamic treatment regimes. Journal of the American Statistical Association, 110:510, 583–598.
Other statistical methods:
earl(),
iqLearn,
optimalClass(),
optimalSeq(),
owl(),
qLearn(),
rwl()
Other weighted learning methods:
earl(),
owl(),
rwl()
Other multiple decision point methods:
iqLearn,
optimalClass(),
optimalSeq(),
qLearn()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # define the negative 4 month change in BMI from baseline y4 <- -100*(bmiData[,5L] - bmiData[,4L])/bmiData[,4L] # reward for second stage rewardSS <- y12 - y4 #### Second-stage regression # Constant propensity model moPropen <- buildModelObj(model = ~1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) fitSS <- bowl(moPropen = moPropen, data = bmiData, reward = rewardSS, txName = 'A2', regime = ~ parentBMI + month4BMI) ##Available methods # Coefficients of the propensity score regression coef(fitSS) # Description of method used to obtain object DTRstep(fitSS) # Estimated value of the optimal treatment regime for training set estimator(fitSS) # Value object returned by propensity score regression method fitObject(fitSS) # Summary of optimization routine optimObj(fitSS) # Estimated optimal treatment for training data optTx(fitSS) # Estimated optimal treatment for new data optTx(fitSS, bmiData) # Plots if defined by propensity regression method dev.new() par(mfrow = c(2,4)) plot(fitSS) plot(fitSS, suppress = TRUE) # Value object returned by propensity score regression method propen(fitSS) # Parameter estimates for decision function regimeCoef(fitSS) # Show main results of method show(fitSS) # Show summary results of method summary(fitSS) #### First-stage regression # Constant propensity model fitFS <- bowl(moPropen = moPropen, data = bmiData, reward = y4, txName = 'A1', regime = ~ gender + parentBMI, BOWLObj = fitSS, lambdas = c(0.5, 1.0), cvFolds = 4L) ##Available methods for fitFS are as shown above for fitSS # Results of the cross-validation cvInfo(fitFS)# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # define the negative 4 month change in BMI from baseline y4 <- -100*(bmiData[,5L] - bmiData[,4L])/bmiData[,4L] # reward for second stage rewardSS <- y12 - y4 #### Second-stage regression # Constant propensity model moPropen <- buildModelObj(model = ~1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) fitSS <- bowl(moPropen = moPropen, data = bmiData, reward = rewardSS, txName = 'A2', regime = ~ parentBMI + month4BMI) ##Available methods # Coefficients of the propensity score regression coef(fitSS) # Description of method used to obtain object DTRstep(fitSS) # Estimated value of the optimal treatment regime for training set estimator(fitSS) # Value object returned by propensity score regression method fitObject(fitSS) # Summary of optimization routine optimObj(fitSS) # Estimated optimal treatment for training data optTx(fitSS) # Estimated optimal treatment for new data optTx(fitSS, bmiData) # Plots if defined by propensity regression method dev.new() par(mfrow = c(2,4)) plot(fitSS) plot(fitSS, suppress = TRUE) # Value object returned by propensity score regression method propen(fitSS) # Parameter estimates for decision function regimeCoef(fitSS) # Show main results of method show(fitSS) # Show summary results of method summary(fitSS) #### First-stage regression # Constant propensity model fitFS <- bowl(moPropen = moPropen, data = bmiData, reward = y4, txName = 'A1', regime = ~ gender + parentBMI, BOWLObj = fitSS, lambdas = c(0.5, 1.0), cvFolds = 4L) ##Available methods for fitFS are as shown above for fitSS # Results of the cross-validation cvInfo(fitFS)
Extends the buildModelObj() function of package modelObj. Here, the returned model object includes a specification of the decision point and subset of the data to which the model is to be applied.
buildModelObjSubset( ..., model, solver.method, solver.args = NULL, predict.method = NULL, predict.args = NULL, dp = 1L, subset = NA )buildModelObjSubset( ..., model, solver.method, solver.args = NULL, predict.method = NULL, predict.args = NULL, dp = 1L, subset = NA )
... |
ignored. Included to require named input. |
model |
An object of class |
solver.method |
An object of class |
solver.args |
An object of class
solver.method = "glm"
solver.args = list("family"=binomial)
See Details section for further information. |
predict.method |
An object of class |
predict.args |
An object of class
predict.method = "predict.glm"
predict.args = list("type"="response").
See Details section for further information. |
dp |
An object of class |
subset |
An object of class |
In some settings, an analyst may want to use different models for unique
subsets of the data. buildModelObjSubset() provides a mechanism for
users to define models for such subset. Specifically, models are specified
in connection with the decision point and subset to which they are to be
applied.
See ?modelObj for further details
An object of class ModelObjSubset, which contains a
complete description of the conditions under which a model is to be
used and the R methods to be used to obtain parameter estimates and
predictions.
# Consider a 2 decision point trial. At the 1st decision point, the subset of # treatment options available to each patient is always set "set1." # At the 2nd decision point, some patients are eligible to receive # treatment from set "set2a" and others from set "set2b." The outcome # for these subsets will be modeled as ~ x1 + x2 and ~ x2 + x3, respectively. # # All parameter estimates are to be obtained used lm and predictions obtained using predict. # # The following illustrates how to build these model objects. model <- list() model[[1]] <- buildModelObjSubset(dp = 1, subset = "set1", model = ~ x1 + x2 + x3, solver.method = 'lm') model[[2]] <- buildModelObjSubset(dp = 2, subset = "set2a", model = ~ ~ x1 + x2, solver.method = 'lm') model[[3]] <- buildModelObjSubset(dp = 2, subset = "set2b", model = ~ x2 + x3, solver.method = 'lm')# Consider a 2 decision point trial. At the 1st decision point, the subset of # treatment options available to each patient is always set "set1." # At the 2nd decision point, some patients are eligible to receive # treatment from set "set2a" and others from set "set2b." The outcome # for these subsets will be modeled as ~ x1 + x2 and ~ x2 + x3, respectively. # # All parameter estimates are to be obtained used lm and predictions obtained using predict. # # The following illustrates how to build these model objects. model <- list() model[[1]] <- buildModelObjSubset(dp = 1, subset = "set1", model = ~ x1 + x2 + x3, solver.method = 'lm') model[[2]] <- buildModelObjSubset(dp = 2, subset = "set2a", model = ~ ~ x1 + x2, solver.method = 'lm') model[[3]] <- buildModelObjSubset(dp = 2, subset = "set2b", model = ~ x2 + x3, solver.method = 'lm')
Returns the unevaluated original call to a DynTxRegime statistical method.
Call(name, ...)Call(name, ...)
name |
Object for which call is desired |
... |
Optional additional input required by R's base call(). |
Methods are defined for all statistical methods implemented in DynTxRegime.
Method retrieves the value object returned by the user specified classification regression modeling object(s). Exact structure of the returned object will vary.
classif(object, ...) ## S4 method for signature 'OptimalClass' classif(object, ...)classif(object, ...) ## S4 method for signature 'OptimalClass' classif(object, ...)
object |
Value object returned from a method that uses classification regression |
... |
Ignored. |
A list is returned, one element for each regression step required by the statistical method.
coef(object, ...)coef(object, ...)
object |
Value object returned by any statistical method implemented in DynTxRegime. |
... |
Optional additional inputs defined by coefficient methods of selected regression functions. |
Methods are defined for all statistical methods implemented in DynTxRegime.
The exact structure of the returned list will vary depending on the statistical method. For methods that include a propensity regression, the returned list will include an element named 'propen'. For methods that include an outcome regression, the returned list will include an element named 'outcome'.
Extract cross-validation results from the value object returned by a weighted learning statistical method of DynTxRegime.
cvInfo(object, ...)cvInfo(object, ...)
object |
A value object returned by a weighted learning statistical method of DynTxRegime |
... |
Ignored. |
Methods are developed for all weighted learning methods implemented in DynTxRegime. Specifically, OWL, RWL, BOWL, and EARL.
Prints are displays a brief description of the statistical method used to obtain the input object.
DTRstep(object)DTRstep(object)
object |
Value object returned by any statistical method of DynTxRegime |
Methods are defined for all statistical methods implemented in DynTxRegime.
Efficient Augmentation and Relaxation Learning
earl( ..., moPropen, moMain, moCont, data, response, txName, regime, iter = 0L, fSet = NULL, lambdas = 0.5, cvFolds = 0L, surrogate = "hinge", kernel = "linear", kparam = NULL, verbose = 2L )earl( ..., moPropen, moMain, moCont, data, response, txName, regime, iter = 0L, fSet = NULL, lambdas = 0.5, cvFolds = 0L, surrogate = "hinge", kernel = "linear", kparam = NULL, verbose = 2L )
... |
Used primarily to require named input. However, inputs for the optimization methods can be sent through the ellipsis. If surrogate is hinge, the optimization method is dfoptim::hjk(). For all other surrogates, stats::optim() is used. |
moPropen |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the propensity for treatment. See ?moPropen for details. |
moMain |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the main effects of the outcome. See ?modelObj for details. |
moCont |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the contrasts of the outcome. See ?modelObj for details. |
data |
A data frame of the covariates and tx histories |
response |
The response variable. |
txName |
A character object. The column header of data that corresponds to the tx covariate |
regime |
A formula object or a list of formula objects. The covariates to be included in classification. If a list is provided, this specifies that there is an underlying subset structure – fSet must then be defined. |
iter |
Maximum number of iterations for outcome regression |
fSet |
A function or NULL defining subset structure |
lambdas |
A numeric object or a numeric vector object giving the penalty tuning parameter. If more than 1 is provided, the finite set of values to be considered in the cross-validation algorithm |
cvFolds |
If cross-validation is to be used to select the tuning parameters, the number of folds. |
surrogate |
The surrogate 0-1 loss function must be one of logit, exp, hinge, sqhinge, huber |
kernel |
A character object. must be one of {"linear", "poly", "radial"} |
kparam |
A numeric object of NULL. If kernel = linear, kparam is ignored. If kernel = poly, kparam is the degree of the polynomial If kernel = radial, kparam is the inverse bandwidth of the kernel. If a vector of bandwidth parameters is given, cross-validation will be used to select the parameter |
verbose |
An integer or logical. If 0, no screen prints are generated. If 1, screen prints are generated with the exception of optimization results obtained in iterative algorithm. If 2, all screen prints are generated. |
an EARL object
Ying-Qi Zhao, Eric Laber, Sumona Saha and Bruce E. Sands (2016+) Efficient augmentation and relaxation learning for treatment regimes using observational data
Other statistical methods:
bowl(),
iqLearn,
optimalClass(),
optimalSeq(),
owl(),
qLearn(),
rwl()
Other single decision point methods:
optimalClass(),
optimalSeq(),
owl(),
qLearn(),
rwl()
Other weighted learning methods:
bowl(),
owl(),
rwl()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') fitEARL <- earl(moPropen = moPropen, moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, surrogate = 'logit', kernel = 'poly', kparam = 2) ##Available methods # Coefficients of the regression objects coef(fitEARL) # Description of method used to obtain object DTRstep(fitEARL) # Estimated value of the optimal treatment regime for training set estimator(fitEARL) # Value object returned by regression methods fitObject(fitEARL) # Summary of optimization routine optimObj(fitEARL) # Estimated optimal treatment for training data optTx(fitEARL) # Estimated optimal treatment for new data optTx(fitEARL, bmiData) # Value object returned by outcome regression method outcome(fitEARL) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(fitEARL) plot(fitEARL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitEARL) # Parameter estimates for decision function regimeCoef(fitEARL) # Show main results of method show(fitEARL) # Show summary results of method summary(fitEARL)# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') fitEARL <- earl(moPropen = moPropen, moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, surrogate = 'logit', kernel = 'poly', kparam = 2) ##Available methods # Coefficients of the regression objects coef(fitEARL) # Description of method used to obtain object DTRstep(fitEARL) # Estimated value of the optimal treatment regime for training set estimator(fitEARL) # Value object returned by regression methods fitObject(fitEARL) # Summary of optimization routine optimObj(fitEARL) # Estimated optimal treatment for training data optTx(fitEARL) # Estimated optimal treatment for new data optTx(fitEARL, bmiData) # Value object returned by outcome regression method outcome(fitEARL) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(fitEARL) plot(fitEARL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitEARL) # Parameter estimates for decision function regimeCoef(fitEARL) # Show main results of method show(fitEARL) # Show summary results of method summary(fitEARL)
EARL
Class EARL contains results for an EARL analysis.
analysisContains a Learning or LearningMulti object.
analysis@txInfoFeasible tx information.
analysis@propenPropensity regression analysis.
analysis@outcomeOutcome regression analysis.
analysis@cvInfoCross-validation analysis if single regime.
analysis@optimOptimization analysis if single regime.
analysis@optimResultlist of cross-validation and optimization results if multiple regimes. optimResult[[i]]@cvInfo and optimResult[[i]]@optim.
analysis@optimalEstimated optimal Tx and value.
analysis@callUnevaluated call to statistical method.
: Retrieve value object returned by outcome regression methods.
: Retrieve value object returned by propensity regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve cross-validation results.
: Retrieve value object returned by optimization method(s).
: Retrieve estimated parameters for optimal tx regime.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Retrieve the value as estimated by the statistical method.
estimator(x, ...) ## S4 method for signature 'IQLearnFS' estimator(x, w = NULL, y = NULL, z = NULL, dens = NULL) ## S4 method for signature 'IQLearnSS' estimator(x, w = NULL, y = NULL, z = NULL, dens = NULL)estimator(x, ...) ## S4 method for signature 'IQLearnFS' estimator(x, w = NULL, y = NULL, z = NULL, dens = NULL) ## S4 method for signature 'IQLearnSS' estimator(x, w = NULL, y = NULL, z = NULL, dens = NULL)
x |
a DynTxRegime Object. |
... |
Optional additional input. Ignored. |
w |
If IQ-Learning, object of class IQLearnSS, IQLearnFS_C, IQLearnFS_ME, or IQLearnFS_VHet |
y |
If IQ-Learning, object of class IQLearnSS, IQLearnFS_C, IQLearnFS_ME, or IQLearnFS_VHet |
z |
If IQ-Learning, object of class IQLearnSS, IQLearnFS_C, IQLearnFS_ME, or IQLearnFS_VHet |
dens |
If IQ-Learning, one of {"norm", "nonpar"} |
Returns a list of the objects returned by all modeling functions
fitObject(object, ...)fitObject(object, ...)
object |
Value object returned by a statistical method of DynTxRegime |
... |
Optional additional inputs |
Methods are defined for all statistical methods implemented in DynTxRegime.
The exact structure of the returned list will vary depending on the statistical method. For methods that include a propensity regression, the returned list will include an element named 'propen'. For methods that include an outcome regression, the returned list will include an element named 'outcome'.
Extracts the contrasts component of the fitted outcome regression the second-stage analysis of the interactive Q-Learning algorithm.
fittedCont(object, ...) ## S4 method for signature 'IQLearnSS' fittedCont(object, ...)fittedCont(object, ...) ## S4 method for signature 'IQLearnSS' fittedCont(object, ...)
object |
An object of class IQLearnSS |
... |
Ignored. |
Extracts the main effects component of the fitted outcome regression for the second-stage analysis of the interactive Q-Learning algorithm.
fittedMain(object, ...) ## S4 method for signature 'IQLearnSS' fittedMain(object, ...)fittedMain(object, ...) ## S4 method for signature 'IQLearnSS' fittedMain(object, ...)
object |
An object of class IQLearnSS |
... |
Ignored. |
Several of the statistical methods implemented in package DynTxRegime allow for subset modeling or limiting of feasible treatment options. This section details how this input is to be defined.
In general, input fSet is used to define subsets of patients
within an analysis. These subsets can be specified to (1) limit
available treatments, (2) use different models for the propensity
score and/or outcome regressions, and/or
(3) use different decision function models for
each subset of patients. The combination of inputs moPropen,
moMain, moCont, fSet, and/or regimes
determines which of these scenarios is
being considered. We cover some common situations below.
Regardless of the purpose for specifying fSet, it must be a
function that returns a list. There are two options for defining the
function. Version 1 is that of the original DynTxRegime package.
In this version, fSet defines the rules
for determining the subset of treatment options for an INDIVIDUAL.
The first element of the returned list is a character, which we term
the subset 'nickname.' This nickname is for bookkeeping purposes
and is used to link models to subsets. The second element
of the returned list is a vector
of available treatment options for the subset. The formal arguments of
the function must include (i) 'data' or (ii) individual covariate
names as given by the column headers of data. An example using the
covariate name input form is
fSet <- function(a1) {
if (a1 > 1) {
subset <- list('subA',c(1,2))
} else {
subset <- list('subB',c(3,4) )
}
return(subset)
}
This function indicates that if an individual has covariate a1 > 1, they are a member of subset 'subA' and their feasible treatment options are {1,2}. If a1 <= 1, they are a member of subset 'subB' and their feasible treatment options are {3,4}.
A more efficient implementation for fSet is now accepted. In
the second form, fSet defines the subset of treatment options
for the full DATASET. It is again a function with
formal arguments (i) 'data' or (ii) individual covariate names as
given by the column headers of data. The function returns a list
containing two elements: 'subsets' and 'txOpts.' Element 'subsets' is
a list comprising all treatment subsets; each element of the list contains
the nickname and treatment options for a single subset. Element
'txOpts' is a character vector indicating the subset of which
each individual is a member. In this new format,
the equivalent definition of fSet as that given above is:
fSet <- function(a1) {
subsets <- list(list('subA', c(1,2)),
list('subB', c(3,4)))
txOpts <- rep('subB', length(x = a1))
txOpts[a1 > 1] <- 'subA'
return(list("subsets" = subsets,
"txOpts" = txOpts))
}
Though a bit more complicated, this version is much more efficient as it processes the entire dataset at once rather than each individual separately.
The simplest scenario involving fSet is to define feasible
treatment options and the rules that dictate how those treatment
options are determined. For example,
responder/non-responder scenarios are often encountered in
multiple-decision-point settings. An example of this scenario is:
patients that respond to the first stage treatment
remain on the original treatment; those that
do not respond to the first stage treatment
have all treatment options available to them at the second stage.
In this case, the
propensity score models for the second stage
are fit using only 'non-responders' for whom
more than 1 treatment option is available.
An example of an appropriate fSet function for
the second-stage is
fSet <- function(data) {
if (data\$responder == 0L) {
subset <- list('subA',c(1L,2L))
} else if (data\$tx1 == 1L) {
subset <- list('subB',c(1L) )
} else if (data\$tx1 == 2L) {
subset <- list('subC',c(2L) )
}
return(subset)
}
for version 1 or for version 2
fSet <- function(data) {
subsets <- list(list('subA', c(1L,2L)),
list('subB', c(1L)),
list('subC', c(2L)))
txOpts <- character(nrow(x = data))
txOpts[data$tx1 == 1L] <- 'subB'
txOpts[data$tx1 == 2L] <- 'subC'
txOpts[data$responder == 0L] <- 'subA'
return(list("subsets" = subsets,
"txOpts" = txOpts))
}
The functions above specify that patients with covariate responder = 0
receive treatments from subset 'subA,' which comprises treatments
A = (1,2). Patients with covariate responder = 1 receive treatment
from subset 'subB' or 'subC' depending on the first stage treatment
received. If
fSet is specified in this way, the form of the model object depends
on the training data. Specifically, if the training data obeys the feasible
treatment rule (here, all individuals with responder = 1 received tx
in accordance with fSet), moPropen would be a "modelObj";
the propensity model will be fit using only those patients with
responder = 0; those with responder = 1 always receive the appropriate
second stage treatment with probability 1.0. However, if the data
are from an observation study and the training data do not obey the
feasible treatment rules (here, some individuals with responder = 1 received
tx = 0; others tx = 1), the responder = 1 data must be modeled and moPropen
must be provided as one or more ModelObjSubset() objects.
If outcome regression is used by the method,
moMain and moCont can be either objects
of class "modelObj" if only responder = 0 patients are to be used
to obtain parameter estimates or as lists of objects of class
"ModelObjSubset" if subsets are to be analyzed individually or
combined for a single fit of all data.
For a scenario where all patients have the same set of treatment
options available, but subsets of patients are to be analyzed using
different models. We cane define fSet as
fSet <- function(data) {
if (data\$a1 == 1) {
subset <- list('subA',c(1L,2L))
} else {
subset <- list('subB',c(1L,2L) )
}
return(subset)
}
for version 1 or in the format of version 2
fSet <- function(data)
{
subsets <- list(list('subA', c(1L,2L)),
list('subB', c(1L,2L)))
txOpts <- rep('subB', nrow(x = data))
txOpts[data$a1 == 1L] <- 'subA'
return(list("subsets" = subsets,
"txOpts" = txOpts))
}
where all patients have the same treatment options available, A = (1,2),
but different regression models will be fit for each subset (case 2 above)
and/or different decision function models (case 3 above) for each
subset. If different propensity score models are used, moPropen
must be a list of objects of class "modelObjSubset."
Perhaps,
propenA <- buildModelObjSubset(model = ~1,
solver.method = 'glm',
solver.args = list('family'='binomial'),
predict.method = 'predict.glm',
predict.args = list(type='response'),
subset = 'subA')
propenB <- buildModelObjSubset(model = ~1,
solver.method = 'glm',
solver.args = list('family'='binomial'),
predict.method = 'predict.glm',
predict.args = list(type='response'),
subset = 'subB')
moPropen <- list(propenA, propenB)
If different decision function models are to be fit, regimes
would take a form similar to
regimes <- list( 'subA' = ~x1 + x2,
'subB' = ~x2 )
Notice that the names of the elements of regimes and the subsets passed to
buildModelObjSubset() correspond to the names defined by fSet,
i.e., 'subA' or 'subB.' These nicknames are used for bookkeeping and
link subsets to the appropriate models.
For a single-decision-point analysis, fSet
is a single function. For multiple-decision-point analyses,
fSet is a list of functions where each element of
the list corresponds to the decision point (1st element <-
1st decision point, etc.)
Retrieve the value object returned by rgenoud() in optimalSeq().
genetic(object, ...) ## S4 method for signature 'OptimalSeq' genetic(object, ...)genetic(object, ...) ## S4 method for signature 'OptimalSeq' genetic(object, ...)
object |
Value object returned by optimalSeq() |
... |
Optional inputs. Ignored. |
The complete interactive Q-Learning algorithm.
## Second-Stage Analysis iqLearnSS(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis for Fitted Main Effects iqLearnFSM(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis for Fitted Contrasts iqLearnFSC(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis of Contrast Variance Log-Linear Model iqLearnFSV(..., object, moMain, moCont, data, iter = 0L, verbose = TRUE)## Second-Stage Analysis iqLearnSS(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis for Fitted Main Effects iqLearnFSM(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis for Fitted Contrasts iqLearnFSC(..., moMain, moCont, data, response, txName, iter = 0L, verbose = TRUE) ## First-Stage Analysis of Contrast Variance Log-Linear Model iqLearnFSV(..., object, moMain, moCont, data, iter = 0L, verbose = TRUE)
... |
ignored. Provided to require named inputs. |
moMain |
An object of class modelObj or a list of objects of class modelObjSubset, which define the models and R methods to be used to obtain parameter estimates and predictions for the main effects component of the outcome regression. See ?modelObj and/or ?modelObjSubset for details. NULL is an acceptable value if moCont is defined. |
moCont |
An object of class modelObj or a list of objects of class modelObjSubset, which define the models and R methods to be used to obtain parameter estimates and predictions for the contrasts component of the outcome regression. See ?modelObj and/or ?modelObjSubset for details. NULL is an acceptable value if moMain is defined. |
data |
A data frame of covariates and treatment history. |
response |
For the second stage analysis, the response vector. For first stage analyses, the value object returned by iqLearnSS(). |
object |
The value object returned by iqLearFSC() |
txName |
A character string giving column header of treatment variable in data |
iter |
An integer. See ?iter for details |
verbose |
A logical. If TRUE, screen prints are generated. |
Laber, EB, Linn, KA, and Stefanski, LA (2014). Interactive model building for Q-Learning. Biometrika, 101, 831–847. PMCID: PMC4274394.
Other statistical methods:
bowl(),
earl(),
optimalClass(),
optimalSeq(),
owl(),
qLearn(),
rwl()
Other multiple decision point methods:
bowl(),
optimalClass(),
optimalSeq(),
qLearn()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] #### Full Interactive Q-Learning Algorithm ### Second-Stage Analysis # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') fitSS <- iqLearnSS(moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2') ### First-Stage Analysis Main Effects Term # main effects model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFSM <- iqLearnFSM(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ### First-Stage Analysis Contrasts Term # contrasts model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFSC <- iqLearnFSC(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ### First-Stage Analysis Contrasts Variance - Log-linear # contrasts variance model moMain <- buildModelObj(model = ~baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~baselineBMI, solver.method = 'lm') fitFSV <- iqLearnFSV(object = fitFSC, moMain = moMain, moCont = moCont, data = bmiData) ####Available methods ### Estimated value estimator(x = fitFSC, y = fitFSM, z = fitFSV, w = fitSS, dens = 'nonpar') ## Estimated optimal treatment and decision functions for training data ## Second stage optimal treatments optTx(x = fitSS) ## First stage optimal treatments when contrast variance is modeled. optTx(x = fitFSM, y = fitFSC, z = fitFSV, dens = 'nonpar') ## First stage optimal treatments when contrast variance is constant. optTx(x = fitFSM, y = fitFSC, dens = 'nonpar') ## Estimated optimal treatment and decision functions for new data ## Second stage optimal treatments optTx(x = fitSS, bmiData) ## First stage optimal treatments when contrast variance is modeled. optTx(x = fitFSM, y = fitFSC, z = fitFSV, dens = 'nonpar', bmiData) ## First stage optimal treatments when contrast variance is constant. optTx(x = fitFSM, y = fitFSC, dens = 'nonpar', bmiData) ### The following methods are available for all objects: fitSS, fitFSM, ### fitFSC and fitFSV. We include only one here for illustration. # Coefficients of the outcome regression objects coef(object = fitSS) # Description of method used to obtain object DTRstep(object = fitFSM) # Value object returned by outcome regression method fitObject(object = fitFSC) # Value object returned by outcome regression method outcome(object = fitFSV) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fitSS) plot(x = fitSS, suppress = TRUE) # Show main results of method show(object = fitFSM) # Show summary results of method summary(object = fitFSV)# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] #### Full Interactive Q-Learning Algorithm ### Second-Stage Analysis # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') fitSS <- iqLearnSS(moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2') ### First-Stage Analysis Main Effects Term # main effects model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFSM <- iqLearnFSM(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ### First-Stage Analysis Contrasts Term # contrasts model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFSC <- iqLearnFSC(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ### First-Stage Analysis Contrasts Variance - Log-linear # contrasts variance model moMain <- buildModelObj(model = ~baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~baselineBMI, solver.method = 'lm') fitFSV <- iqLearnFSV(object = fitFSC, moMain = moMain, moCont = moCont, data = bmiData) ####Available methods ### Estimated value estimator(x = fitFSC, y = fitFSM, z = fitFSV, w = fitSS, dens = 'nonpar') ## Estimated optimal treatment and decision functions for training data ## Second stage optimal treatments optTx(x = fitSS) ## First stage optimal treatments when contrast variance is modeled. optTx(x = fitFSM, y = fitFSC, z = fitFSV, dens = 'nonpar') ## First stage optimal treatments when contrast variance is constant. optTx(x = fitFSM, y = fitFSC, dens = 'nonpar') ## Estimated optimal treatment and decision functions for new data ## Second stage optimal treatments optTx(x = fitSS, bmiData) ## First stage optimal treatments when contrast variance is modeled. optTx(x = fitFSM, y = fitFSC, z = fitFSV, dens = 'nonpar', bmiData) ## First stage optimal treatments when contrast variance is constant. optTx(x = fitFSM, y = fitFSC, dens = 'nonpar', bmiData) ### The following methods are available for all objects: fitSS, fitFSM, ### fitFSC and fitFSV. We include only one here for illustration. # Coefficients of the outcome regression objects coef(object = fitSS) # Description of method used to obtain object DTRstep(object = fitFSM) # Value object returned by outcome regression method fitObject(object = fitFSC) # Value object returned by outcome regression method outcome(object = fitFSV) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fitSS) plot(x = fitSS, suppress = TRUE) # Show main results of method show(object = fitFSM) # Show summary results of method summary(object = fitFSV)
IQLearnFS_C
Class IQLearnFS_C contains the results for the first stage
contrasts component of the interactive Q-Learning algorithm.
Objects of this class are returned by iqLearnFSC().
txVec: A numeric. treatment vector from training data
residuals: A numeric. residuals of the fit
step: Not used in this context.
outcome: The outcome regression analysis
txInfo: The feasible tx information
optimal: The estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
:Retrieve the residuals of the regression.
:Retrieve the standard deviation of the residuals.
IQLearnFS_ME
Class IQLearnFS_ME contains the results for the first stage
main effects component of the interactive Q-Learning algorithm.
Objects of this class are returned by iqLearnFSM().
step: Not used in this context.
outcome: The outcome regression analysis
txInfo: The feasible tx information
optimal: The estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
IQLearnFS_VHet
Class IQLearnFS_VHet contains the results for the first stage
residuals component of the interactive Q-Learning algorithm.
Objects of this class are returned by iqLearnFSV().
residuals: Standardized residuals of contrast after modeling
scale: Scaling factor for stdization
step: Not used in this context.
outcome: The outcome regression analysis
txInfo: The feasible tx information
optimal: The estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
:Retrieve the residuals of the regression.
QQ plot of the residuals for the interactive Q-Learning algorithm.
IQLearnSS
Class IQLearnSS contains the results for the second stage
of the interactive Q-Learning algorithm. Objects of this class are
returned by iqLearnSS().
yContHat: A numeric. Estimated contrast component
yMainHat: A numeric. Estimated main effects component
delta: A numeric. Indicator of compliance * response used for value calc
step: Not used in this context.
outcome: The outcome regression analysis
txInfo: The feasible tx information
optimal: The estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
:Retrieve the contrasts component of the regression.
:Retrieve the main effects component of the regression.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Several of the statistical methods implemented in package DynTxRegime allow for an iterative algorithm when completing an outcome regression. This section details how this input is to be defined.
Outcome regression models are specified by the main effects components
(moMain) and the contrasts component (moCont).
Assuming that the
treatment is denoted as binary A, the full regression model is:
moMain + A*moCont. There are two ways to fit this model: (i)
in the full model formulation (moMain + A*moCont) or (ii) each
component, moMain and moCont, is fit separately.
iter specifies
if (i) or (ii) should be used.
iter >= 1 indicates that moMain and moCont
are to be
fit separately using an iterative algorithm.
iter is the maximum number of iterations.
Assume Y = Ymain + Ycont;
the iterative algorithm is as follows:
(1) hat(Ycont) = 0;
(2) Ymain = Y - hat(Ycont);
(3) fit Ymain ~ moMain;
(4) set Ycont = Y - hat(Ymain)
(5) fit Ycont ~ A*moCont;
(6) Repeat steps (2) - (5) until convergence or a maximum of iter iterations.
This choice allows the user to specify, for example, a linear main effects component and a non-linear contrasts component.
iter <= 0 indicates that the full model formulation is to be
used. The components moMain and moCont will be
combined in the package and fit as a single object.
Note that if iter <= 0, all non-model components of
moMain and moCont must be identical. Specifically,
the regression method and any non-default arguments
should be identical.
By default, the specifications in moMain are used.
Several of the statistical methods implemented in package DynTxRegime use propensity score modeling. This section details how this input is to be defined.
For input moPropen, the method specified to obtain predictions
MUST return the prediction on the scale of the probability,
i.e., predictions must be in the range (0,1). In
addition, moPropen differs from standard "modelObj"
objects in that an additional element may be required in
predict.args. Recall, predict.args is the list of control
parameters passed to the prediction method. An additional control
parameter, propen.missing can be included. propen.missing
takes value "smallest" or "largest". It will be required if the
prediction method returns predictions for only a subset of the
treatment data; e.g., predict.glm(). propen.missing indicates if
it is the smallest or the largest treatment value that is missing
from the returned predictions.
For example, fitting a binary treatment (A in {0,1}) using
moPropen <- buildModelObj(model = ~1,
solver.method = 'glm',
solver.args = list('family'='binomial'),
predict.method = 'predict.glm',
predict.args = list(type='response'))
returns only P(A=1). P(A=0) is "missing," and thus
moPropen <- buildModelObj(model = ~1,
solver.method = 'glm',
solver.args = list('family'='binomial'),
predict.method = 'predict.glm',
predict.args = list(type='response',
propen.missing = 'smallest'))
If the dimension of the value returned by the prediction method is
less than the number of treatment options and no value is provided
in propen.missing, it is assumed that the smallest valued treatment
option is missing. Here, 'smallest' indicates the lowest value
integer if treatment is an integer, or the 'base' level if treatment
is a factor.
Classification Perspective
optimalClass( ..., moPropen, moMain, moCont, moClass, data, response, txName, iter = 0L, fSet = NULL, verbose = TRUE )optimalClass( ..., moPropen, moMain, moCont, moClass, data, response, txName, iter = 0L, fSet = NULL, verbose = TRUE )
... |
Included to require named inputs |
moPropen |
An object of class modelObj, which defines the models and R methods to be used to obtain parameter estimates and predictions for the propensity for treatment. See ?moPropen for details. |
moMain |
An object of class modelObj, which defines the models and R methods to be used to obtain parameter estimates and predictions for for the main effects component of the outcome regression. See ?modelObj for details. NULL is an appropriate value. |
moCont |
An object of class modelObj, which defines the models and R methods to be used to obtain parameter estimates and predictions for for the contrasts component of the outcome regression. See ?modelObj for details. NULL is an appropriate value. |
moClass |
An object of class modelObj, which defines the models and R methods to be used to obtain parameter estimates and predictions for the classification. See ?modelObj for details. |
data |
A data frame of the covariates and tx histories |
response |
The response vector |
txName |
An character giving the column header of the column in data that contains the tx covariate. |
iter |
An integer See ?iter for details |
fSet |
A function or NULL. This argument allows the user to specify the subset of tx options available to a patient. See ?fSet for details of allowed structure |
verbose |
A logical If FALSE, screen prints are suppressed. |
an object of class OptimalClass
Baqun Zhang, Anastasios A. Tsiatis, Marie Davidian, Min Zhang and Eric B. Laber. "Estimating optimal tx regimes from a classification perspective." Stat 2012; 1: 103-114.
Note that this method is a single decision point, binary treatment method. For multiple decision points, can be called repeatedly.
Other statistical methods:
bowl(),
earl(),
iqLearn,
optimalSeq(),
owl(),
qLearn(),
rwl()
Other single decision point methods:
earl(),
optimalSeq(),
owl(),
qLearn(),
rwl()
Other multiple decision point methods:
bowl(),
iqLearn,
optimalSeq(),
qLearn()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # classification model library(rpart) moClass <- buildModelObj(model = ~parentBMI+month4BMI+race+gender, solver.method = 'rpart', solver.args = list(method="class"), predict.args = list(type='class')) #### Second-Stage Analysis using IPW fitSS_IPW <- optimalClass(moPropen = moPropen, moClass = moClass, data = bmiData, response = y12, txName = 'A2') # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') #### Second-Stage Analysis using AIPW fitSS_AIPW <- optimalClass(moPropen = moPropen, moMain = moMain, moCont = moCont, moClass = moClass, data = bmiData, response = y12, txName = 'A2') ##Available methods # Retrieve the classification regression object classif(object = fitSS_AIPW) # Coefficients of the outcome regression objects coef(object = fitSS_AIPW) # Description of method used to obtain object DTRstep(object = fitSS_AIPW) # Estimated value of the optimal treatment regime for training set estimator(x = fitSS_AIPW) # Value object returned by outcome regression method fitObject(object = fitSS_AIPW) # Estimated optimal treatment and decision functions for training data optTx(x = fitSS_AIPW) # Estimated optimal treatment and decision functions for new data optTx(x = fitSS_AIPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fitSS_AIPW) outcome(object = fitSS_IPW) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fitSS_AIPW) plot(x = fitSS_AIPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fitSS_AIPW) # Show main results of method show(object = fitSS_AIPW) # Show summary results of method summary(object = fitSS_AIPW) #### First-stage Analysis using AIPW # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # classification model moClass <- buildModelObj(model = ~parentBMI+baselineBMI+race+gender, solver.method = 'rpart', solver.args = list(method="class"), predict.args = list(type='class')) # outcome model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFS_AIPW <- optimalClass(moPropen = moPropen, moMain = moMain, moCont = moCont, moClass = moClass, data = bmiData, response = fitSS_AIPW, txName = 'A1') ##Available methods for fitFS_AIPW are as shown above for fitSS_AIPW# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # classification model library(rpart) moClass <- buildModelObj(model = ~parentBMI+month4BMI+race+gender, solver.method = 'rpart', solver.args = list(method="class"), predict.args = list(type='class')) #### Second-Stage Analysis using IPW fitSS_IPW <- optimalClass(moPropen = moPropen, moClass = moClass, data = bmiData, response = y12, txName = 'A2') # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') #### Second-Stage Analysis using AIPW fitSS_AIPW <- optimalClass(moPropen = moPropen, moMain = moMain, moCont = moCont, moClass = moClass, data = bmiData, response = y12, txName = 'A2') ##Available methods # Retrieve the classification regression object classif(object = fitSS_AIPW) # Coefficients of the outcome regression objects coef(object = fitSS_AIPW) # Description of method used to obtain object DTRstep(object = fitSS_AIPW) # Estimated value of the optimal treatment regime for training set estimator(x = fitSS_AIPW) # Value object returned by outcome regression method fitObject(object = fitSS_AIPW) # Estimated optimal treatment and decision functions for training data optTx(x = fitSS_AIPW) # Estimated optimal treatment and decision functions for new data optTx(x = fitSS_AIPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fitSS_AIPW) outcome(object = fitSS_IPW) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fitSS_AIPW) plot(x = fitSS_AIPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fitSS_AIPW) # Show main results of method show(object = fitSS_AIPW) # Show summary results of method summary(object = fitSS_AIPW) #### First-stage Analysis using AIPW # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # classification model moClass <- buildModelObj(model = ~parentBMI+baselineBMI+race+gender, solver.method = 'rpart', solver.args = list(method="class"), predict.args = list(type='class')) # outcome model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFS_AIPW <- optimalClass(moPropen = moPropen, moMain = moMain, moCont = moCont, moClass = moClass, data = bmiData, response = fitSS_AIPW, txName = 'A1') ##Available methods for fitFS_AIPW are as shown above for fitSS_AIPW
OptimalClass
Class OptimalClass contains results for a single decision point
when estimates are obtained from the classification perspective.
Objects of this class are returned by optimalClass().
stepStep in the algorithm.
analysisAnalysis results.
: Retrieve value object returned by outcome regression methods.
: Retrieve value object returned by propensity regression methods.
: Retrieve value object returned by classification regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
OptimalClassObj
Class OptimalClassObj contains results for a single decision point
when estimates are obtained from the classification perspective.
Objects of this class are returned by optimalClass().
classResults of the classification step.
outcomeResults of the outcome regression step.
propenResults of the propensity step.
optimalEstimated optimal tx and value
CallUnevaluated call.
: Retrieve value object returned by outcome regression methods.
: Retrieve value object returned by propensity regression methods.
: Retrieve value object returned by classification regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
OptimalInfo
Class OptimalInfo stores the estimated optimal tx, decision functions,
and estimated value.
optimalTxa vector of the estimated optimal tx
estimatedValuea vector of the estimated value
decisionFunca vector or matrix containing the values used to determine @optimalTx (if applicable)
Missing or Coarsened Data Perspective - Genetic Algorithm
optimalSeq( ..., moPropen, moMain, moCont, data, response, txName, regimes, fSet = NULL, refit = FALSE, iter = 0L, verbose = TRUE )optimalSeq( ..., moPropen, moMain, moCont, data, response, txName, regimes, fSet = NULL, refit = FALSE, iter = 0L, verbose = TRUE )
... |
Additional arguments required by rgenoud. At a minimum this should include Domains, pop.size and starting.values. See ?rgenoud for more information. |
moPropen |
An object of class modelObj, a list of objects of class modelObj, or a list of object of class modelObjSubset, which define the models and R methods to be used to obtain parameter estimates and predictions for the propensity for treatment. See ?moPropen for details. |
moMain |
An object of class modelObj, a list of objects of class modelObj, or a list of object of class modelObjSubset, which define the models and R methods to be used to obtain parameter estimates and predictions for the main effects component of the outcome regression. See ?modelObj and/or ?modelObjSubset for details. NULL is an acceptable input if IPWE is desired or there is no main effects component of the outcome regression model. |
moCont |
An object of class modelObj, a list of objects of class modelObj, or a list of object of class modelObjSubset, which define the models and R methods to be used to obtain parameter estimates and predictions for the contrasts component of the outcome regression. See ?modelObj and/or ?modelObjSubset for details. NULL is an acceptable input if IPWE is desired or there is no contrast component of the outcome regression model. |
data |
A data frame of the covariates and tx history |
response |
The response vector |
txName |
A vector of characters. The column headers of data that correspond to the tx covariate for each decision point. The ordering should be sequential, i.e., the 1st element gives column name for the 1st decision point tx, the 2nd gives column name for the 2nd decision point tx, etc. |
regimes |
A function or a list of functions. For each decision point, a function defining the tx rule. For example, if the tx rule is : I(eta_1 < x1), regimes is defined as regimes <- function(a,data) {as.numeric(a < data$x1)} THE LAST ARGUMENT IS ALWAYS TAKEN TO BE THE DATA.FRAME |
fSet |
A function or a list of functions. This argument allows the user to specify the subset of tx options available to a patient or the subset of patients that will be modeled uniquely. see ?fSet for details |
refit |
No longer used |
iter |
An integer. See ?iter for details |
verbose |
A logical. If FALSE, screen prints are suppressed. |
An object inheriting from class OptimalSeq
Baqun Zhang, Anastasios A. Tsiatis, Eric B. Laber & Marie Davidian, "A Robust Method for Estimating Optimal Treatment Regimes", Biometrics, 68, 1010-1018.
Baqun Zhang, Anastasios A. Tsiatis, Eric B. Laber & Marie Davidian, "Robust estimation of optimal treatment regimes for sequential treatment decisions", Biometrika (2013) pp.1-14.
Other statistical methods:
bowl(),
earl(),
iqLearn,
optimalClass(),
owl(),
qLearn(),
rwl()
Other single decision point methods:
earl(),
optimalClass(),
owl(),
qLearn(),
rwl()
Other multiple decision point methods:
bowl(),
iqLearn,
optimalClass(),
qLearn()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # Define the propensity for treatment model and methods. # Will use constant model for both decision points moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) moPropen <- list(moPropen, moPropen) # outcome model second stage moMain2 <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont2 <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') # outcome model first stage moMain1 <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont1 <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') moMain <- list(moMain1, moMain2) moCont <- list(moCont1, moCont2) # regime function second stage regime2 <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$month4BMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } # regime function first stage regime1 <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$baselineBMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } regimes <- list(regime1, regime2) #### Analysis using AIPW ## Not run: fit_AIPW <- optimalSeq(moPropen = moPropen, moMain = moMain, moCont = moCont, regimes = regimes, data = bmiData, response = y12, txName = c('A1','A2'), Domains = cbind(rep(0,4),rep(100,4)), pop.size = 100, starting.values = rep(25,4)) ##Available methods # Coefficients of the regression objects coef(object = fit_AIPW) # Description of method used to obtain object DTRstep(object = fit_AIPW) # Estimated value of the optimal treatment regime for training set estimator(x = fit_AIPW) # Value object returned by regression methods fitObject(object = fit_AIPW) # Retrieve the results of genetic algorithm genetic(object = fit_AIPW) # Estimated optimal treatment and decision functions for training data optTx(x = fit_AIPW) # Estimated optimal treatment and decision functions for new data optTx(x = fit_AIPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fit_AIPW) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(x = fit_AIPW) plot(x = fit_AIPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fit_AIPW) # Show main results of method show(object = fit_AIPW) # Show summary results of method summary(object = fit_AIPW) ## End(Not run) #### Single Decision Point Analysis using IPW # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # regime function second stage regime <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$month4BMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } ## Not run: fit_IPW <- optimalSeq(moPropen = moPropen, regimes = regime, data = bmiData, response = y12, txName = 'A2', Domains = cbind(rep(0,2),rep(100,2)), pop.size = 100, starting.values = rep(25,2)) ##Available methods # Coefficients of the regression objects coef(object = fit_IPW) # Description of method used to obtain object DTRstep(object = fit_IPW) # Estimated value of the optimal treatment regime for training set estimator(x = fit_IPW) # Value object returned by regression method fitObject(object = fit_IPW) # Retrieve the results of genetic algorithm genetic(object = fit_IPW) # Estimated optimal treatment and decision functions for training data optTx(x = fit_IPW) # Estimated optimal treatment and decision functions for new data optTx(x = fit_IPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fit_IPW) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fit_IPW) plot(x = fit_IPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fit_IPW) # Show main results of method show(object = fit_IPW) # Show summary results of method summary(object = fit_IPW) ## End(Not run)# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # Define the propensity for treatment model and methods. # Will use constant model for both decision points moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) moPropen <- list(moPropen, moPropen) # outcome model second stage moMain2 <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont2 <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') # outcome model first stage moMain1 <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont1 <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') moMain <- list(moMain1, moMain2) moCont <- list(moCont1, moCont2) # regime function second stage regime2 <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$month4BMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } # regime function first stage regime1 <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$baselineBMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } regimes <- list(regime1, regime2) #### Analysis using AIPW ## Not run: fit_AIPW <- optimalSeq(moPropen = moPropen, moMain = moMain, moCont = moCont, regimes = regimes, data = bmiData, response = y12, txName = c('A1','A2'), Domains = cbind(rep(0,4),rep(100,4)), pop.size = 100, starting.values = rep(25,4)) ##Available methods # Coefficients of the regression objects coef(object = fit_AIPW) # Description of method used to obtain object DTRstep(object = fit_AIPW) # Estimated value of the optimal treatment regime for training set estimator(x = fit_AIPW) # Value object returned by regression methods fitObject(object = fit_AIPW) # Retrieve the results of genetic algorithm genetic(object = fit_AIPW) # Estimated optimal treatment and decision functions for training data optTx(x = fit_AIPW) # Estimated optimal treatment and decision functions for new data optTx(x = fit_AIPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fit_AIPW) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(x = fit_AIPW) plot(x = fit_AIPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fit_AIPW) # Show main results of method show(object = fit_AIPW) # Show summary results of method summary(object = fit_AIPW) ## End(Not run) #### Single Decision Point Analysis using IPW # Define the propensity for treatment model and methods. moPropen <- buildModelObj(model = ~ 1, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # regime function second stage regime <- function(eta1, eta2, data) { tst <- {data$parentBMI > eta1} & {data$month4BMI > eta2} rec <- rep('MR', nrow(x = data)) rec[!tst] <- 'CD' return( rec ) } ## Not run: fit_IPW <- optimalSeq(moPropen = moPropen, regimes = regime, data = bmiData, response = y12, txName = 'A2', Domains = cbind(rep(0,2),rep(100,2)), pop.size = 100, starting.values = rep(25,2)) ##Available methods # Coefficients of the regression objects coef(object = fit_IPW) # Description of method used to obtain object DTRstep(object = fit_IPW) # Estimated value of the optimal treatment regime for training set estimator(x = fit_IPW) # Value object returned by regression method fitObject(object = fit_IPW) # Retrieve the results of genetic algorithm genetic(object = fit_IPW) # Estimated optimal treatment and decision functions for training data optTx(x = fit_IPW) # Estimated optimal treatment and decision functions for new data optTx(x = fit_IPW, newdata = bmiData) # Value object returned by outcome regression method outcome(object = fit_IPW) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(x = fit_IPW) plot(x = fit_IPW, suppress = TRUE) # Retrieve the value object returned by propensity regression method propen(object = fit_IPW) # Show main results of method show(object = fit_IPW) # Show summary results of method summary(object = fit_IPW) ## End(Not run)
OptimalSeq
Class OptimalSeq contains the results for the estimated optimal tx
and value when estimated from a coarsened or missing data perspective.
geneticA list containing the results from the genetic algorithm
propenResults of the propensity regression step
outcomeResults of the outcome regression step
regimeResults for the regime.
optimalResults for the estimated optimal tx and value
CallThe unevaluated call.
: Retrieve value object returned by outcome regression methods.
: Retrieve value object returned by propensity regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve the estimated regime parameters.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Methods for multiple decision point analyses. Class inherits directly from
OptimalSeq and all methods defined for objects of class OptimaSeq
are defined for this class.
Methods for single decision point analyses. Class inherits directly from
OptimalSeq and all methods defined for objects of class OptimaSeq
are defined for this class.
Retrieves the value object returned by the optimization method for weighted learning methods.
optimObj(object, ...) ## S4 method for signature 'OWL' optimObj(object, ...) ## S4 method for signature 'RWL' optimObj(object, ...) ## S4 method for signature 'BOWL' optimObj(object, ...) ## S4 method for signature 'EARL' optimObj(object, ...)optimObj(object, ...) ## S4 method for signature 'OWL' optimObj(object, ...) ## S4 method for signature 'RWL' optimObj(object, ...) ## S4 method for signature 'BOWL' optimObj(object, ...) ## S4 method for signature 'EARL' optimObj(object, ...)
object |
A value object returned by a statistical method of DynTxRegime that uses optimization to estimate regime parameters. |
... |
Ignored. |
If newdata is provided, the results of the statistical method are used to estimate the decision functions and/or optimal tx. If newdata is missing, the estimated decision functions and/or optimal tx obtained for the original training data are returned.
optTx(x, newdata, ...) ## S4 method for signature 'IQLearnFS,data.frame' optTx(x, newdata, ..., y = NULL, z = NULL, dens = NULL) ## S4 method for signature 'IQLearnFS,missing' optTx(x, newdata, ..., y = NULL, z = NULL, dens = NULL)optTx(x, newdata, ...) ## S4 method for signature 'IQLearnFS,data.frame' optTx(x, newdata, ..., y = NULL, z = NULL, dens = NULL) ## S4 method for signature 'IQLearnFS,missing' optTx(x, newdata, ..., y = NULL, z = NULL, dens = NULL)
x |
a DynTxRegime Object. |
newdata |
Optional data.frame if estimates for new patients are desired. |
... |
Optional additional input. |
y |
Object of class IQLearnFS |
z |
Object of class IQLearnFS |
dens |
one of {norm, nonpar} |
Methods are defined for all statistical methods implemented in DynTxRegime.
For statistical methods that require an outcome regression analysis, the value object returned by the modeling function(s) is retrieved.
outcome(object, ...)outcome(object, ...)
object |
A value object returned by a statistical method of DynTxRegime. |
... |
Ignored. |
Methods are defined for all statistical methods implemented in DynTxRegime that use outcome regression.
Outcome Weighted Learning
owl( ..., moPropen, data, reward, txName, regime, response, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, surrogate = "hinge", verbose = 2L )owl( ..., moPropen, data, reward, txName, regime, response, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, surrogate = "hinge", verbose = 2L )
... |
Used primarily to require named input. However, inputs for the optimization methods can be sent through the ellipsis. If surrogate is hinge, the optimization method is kernlab::ipop(). For all other surrogates, stats::optim() is used. |
moPropen |
An object of class modelObj, which defines the model and R methods to be used to obtain parameter estimates and predictions for the propensity for treatment. See ?moPropen for details. |
data |
A data frame of the covariates and tx histories |
reward |
The response vector |
txName |
A character object. The column header of data that corresponds to the tx covariate |
regime |
A formula object or a character vector. The covariates to be included in classification |
response |
A numeric vector. The reward. Allows for naming convention followed in most DynTxRegime methods. |
lambdas |
A numeric object or a numeric vector object giving the penalty tuning parameter. If more than 1 is provided, the finite set of values to be considered in the cross-validation algorithm |
cvFolds |
If cross-validation is to be used to select the tuning parameters, the number of folds. |
kernel |
A character object. must be one of {"linear", "poly", "radial"} |
kparam |
A numeric object of NULL. If kernel = linear, kparam is ignored. If kernel = poly, kparam is the degree of the polynomial If kernel = radial, kparam is the inverse bandwidth of the kernel. If a vector of bandwidth parameters is given, cross-validation will be used to select the parameter |
surrogate |
The surrogate 0-1 loss function must be one of logit, exp, hinge, sqhinge, huber |
verbose |
An integer or logical. If 0, no screen prints are generated. If 1, screen prints are generated with the exception of optimization results obtained in iterative algorithm. If 2, all screen prints are generated. |
an OWL object
Yingqi Zhao, Donglin Zeng, A. John Rush, Michael R. Kosorok (2012) Estimated individualized treatment rules using outcome weighted learning. Journal of the American Statistical Association, 107(409): 1106-1118. PMCID: 3636816
Other statistical methods:
bowl(),
earl(),
iqLearn,
optimalClass(),
optimalSeq(),
qLearn(),
rwl()
Other weighted learning methods:
bowl(),
earl(),
rwl()
Other single decision point methods:
earl(),
optimalClass(),
optimalSeq(),
qLearn(),
rwl()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) fitOWL <- owl(moPropen = moPropen, data = bmiData, reward = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, surrogate = 'hinge', kernel = 'linear', kparam = NULL) ##Available methods # Coefficients of the propensity score regression coef(fitOWL) # Description of method used to obtain object DTRstep(fitOWL) # Estimated value of the optimal treatment regime for training set estimator(fitOWL) # Value object returned by propensity score regression method fitObject(fitOWL) # Summary of optimization routine optimObj(fitOWL) # Estimated optimal treatment for training data optTx(fitOWL) # Estimated optimal treatment for new data optTx(fitOWL, bmiData) # Plots if defined by propensity regression method dev.new() par(mfrow = c(2,4)) plot(fitOWL) plot(fitOWL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitOWL) # Parameter estimates for decision function regimeCoef(fitOWL) # Show main results of method show(fitOWL) # Show summary results of method summary(fitOWL)# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) fitOWL <- owl(moPropen = moPropen, data = bmiData, reward = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, surrogate = 'hinge', kernel = 'linear', kparam = NULL) ##Available methods # Coefficients of the propensity score regression coef(fitOWL) # Description of method used to obtain object DTRstep(fitOWL) # Estimated value of the optimal treatment regime for training set estimator(fitOWL) # Value object returned by propensity score regression method fitObject(fitOWL) # Summary of optimization routine optimObj(fitOWL) # Estimated optimal treatment for training data optTx(fitOWL) # Estimated optimal treatment for new data optTx(fitOWL, bmiData) # Plots if defined by propensity regression method dev.new() par(mfrow = c(2,4)) plot(fitOWL) plot(fitOWL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitOWL) # Parameter estimates for decision function regimeCoef(fitOWL) # Show main results of method show(fitOWL) # Show summary results of method summary(fitOWL)
OWL
Class OWL contains results for an OWL analysis.
analysisContains a Learning or LearningMulti object.
analysis@txInfoFeasible tx information.
analysis@propenPropensity regression analysis.
analysis@outcomeOutcome regression analysis.
analysis@cvInfoCross-validation analysis if single regime.
analysis@optimOptimization analysis if single regime.
analysis@optimResultlist of cross-validation and optimization results if multiple regimes. optimResult[[i]]@cvInfo and optimResult[[i]]@optim.
analysis@optimalEstimated optimal Tx and value.
analysis@callUnevaluated call to statistical method.
: Retrieve value object returned by propensity regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve cross-validation results.
: Retrieve value object returned by optimization method(s).
: Retrieve estimated parameters for optimal tx regime.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Calls plot() method for all regression steps of a statistical method
x |
Value object returned by a statistical method |
y |
Ignored |
suppress |
T/F indicating if titles should be concatenated with information indicating the specific regression step |
... |
Optional additional inputs |
Methods are defined for all statistical methods implemented in DynTxRegime.
For statistical methods that require a propensity regression analysis, the value object returned by the modeling function(s) is retrieved.
propen(object, ...)propen(object, ...)
object |
A value object returned by a statistical method of DynTxRegime. |
... |
Ignored. |
Methods are defined for all statistical methods implemented in DynTxRegime that use propensity regression.
Performs a single step of the Q-Learning algorithm.
If an object of class QLearn is passed through input response,
it is assumed that the QLearn object is the value object returned
from the preceding step of the Q-Learning algorithm, and
the value fit by the regression is taken from the QLearn object.
If a vector is passed through input response, it is assumed that the
call if for the first step in the Q-Learning algorithm, and
models are fit using the provided response.
qLearn( ..., moMain, moCont, data, response, txName, fSet = NULL, iter = 0L, verbose = TRUE )qLearn( ..., moMain, moCont, data, response, txName, fSet = NULL, iter = 0L, verbose = TRUE )
... |
ignored. Provided to require named inputs. |
moMain |
An object of class modelObj or a list of objects of class
modelObjSubset, which define the models and R methods to be used to
obtain parameter estimates and predictions for the main effects component
of the outcome regression. |
moCont |
An object of class modelObj or a list of objects of class
modelObjSubset, which define the models and R methods to be used to
obtain parameter estimates and predictions for the contrasts component
of the outcome regression. |
data |
A data frame of covariates and treatment history. |
response |
A response vector or object of class QLearn from a previous Q-Learning step. |
txName |
A character string giving column header of treatment variable in data |
fSet |
NULL or a function. This argument allows the user to specify the subset of treatment options available to a patient. See ?fSet for details of allowed structure |
iter |
An integer. See ?iter for details |
verbose |
A logical. If TRUE, screen prints are generated. |
An object of class QLearn-class
Other statistical methods:
bowl(),
earl(),
iqLearn,
optimalClass(),
optimalSeq(),
owl(),
rwl()
Other multiple decision point methods:
bowl(),
iqLearn,
optimalClass(),
optimalSeq()
Other single decision point methods:
earl(),
optimalClass(),
optimalSeq(),
owl(),
rwl()
# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') #### Second-Stage Analysis fitSS <- qLearn(moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2') ##Available methods # Coefficients of the outcome regression objects coef(fitSS) # Description of method used to obtain object DTRstep(fitSS) # Estimated value of the optimal treatment regime for training set estimator(fitSS) # Value object returned by outcome regression method fitObject(fitSS) # Estimated optimal treatment and decision functions for training data optTx(fitSS) # Estimated optimal treatment and decision functions for new data optTx(fitSS, bmiData) # Value object returned by outcome regression method outcome(fitSS) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(fitSS) plot(fitSS, suppress = TRUE) # Show main results of method show(fitSS) # Show summary results of method summary(fitSS) #### First-stage Analysis # outcome model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFS <- qLearn(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ##Available methods for fitFS are as shown above for fitSS# Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+month4BMI, solver.method = 'lm') #### Second-Stage Analysis fitSS <- qLearn(moMain = moMain, moCont = moCont, data = bmiData, response = y12, txName = 'A2') ##Available methods # Coefficients of the outcome regression objects coef(fitSS) # Description of method used to obtain object DTRstep(fitSS) # Estimated value of the optimal treatment regime for training set estimator(fitSS) # Value object returned by outcome regression method fitObject(fitSS) # Estimated optimal treatment and decision functions for training data optTx(fitSS) # Estimated optimal treatment and decision functions for new data optTx(fitSS, bmiData) # Value object returned by outcome regression method outcome(fitSS) # Plots if defined by outcome regression method dev.new() par(mfrow = c(2,4)) plot(fitSS) plot(fitSS, suppress = TRUE) # Show main results of method show(fitSS) # Show summary results of method summary(fitSS) #### First-stage Analysis # outcome model moMain <- buildModelObj(model = ~parentBMI+baselineBMI, solver.method = 'lm') moCont <- buildModelObj(model = ~race + parentBMI+baselineBMI, solver.method = 'lm') fitFS <- qLearn(moMain = moMain, moCont = moCont, data = bmiData, response = fitSS, txName = 'A1') ##Available methods for fitFS are as shown above for fitSS
QLearn
Class QLearn contains the results for a Q-Learning step
stepAn integer indicating the step of the Q-Learning algorithm.
outcomeThe outcome regression analysis
txInfoThe feasible tx information
optimalThe estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
QLearnObj
Class QLearnObj contains the results for a Q-Learning step
outcomeThe outcome regression analysis
txInfoThe feasible tx information
optimalThe estimated optimal tx, decision function, and value
: Retrieve value object returned by outcome regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Extract the estimated regime parameters.
regimeCoef(object, ...)regimeCoef(object, ...)
object |
A value object returned by a statistical method of DynTxRegime. |
... |
Ignored. |
Methods are defined for all statistical methods implemented in DynTxRegime that use a non-regression based regime. Specifically, OptimalSeq, OWL, BOWL, RWL, and EARL.
Retrieve residuals from an interactive Q-Learning step.
residuals(object, ...) ## S4 method for signature 'IQLearnFS_C' residuals(object, ...) ## S4 method for signature 'IQLearnFS_VHet' residuals(object, ...)residuals(object, ...) ## S4 method for signature 'IQLearnFS_C' residuals(object, ...) ## S4 method for signature 'IQLearnFS_VHet' residuals(object, ...)
object |
A value object returned by iqLearnC() or iqLearnVar() |
... |
Ignored. |
Residual Weighted Learning
rwl( ..., moPropen, moMain, data, reward, txName, regime, response, fSet = NULL, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, responseType = "continuous", verbose = 2L )rwl( ..., moPropen, moMain, data, reward, txName, regime, response, fSet = NULL, lambdas = 2, cvFolds = 0L, kernel = "linear", kparam = NULL, responseType = "continuous", verbose = 2L )
... |
Used primarily to require named input. However, inputs for the optimization methods can be sent through the ellipsis. The optimization method is stats::optim(). |
moPropen |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the propensity for treatment. See ?moPropen for details. |
moMain |
An object of class modelObj or modelObjSubset, which defines the model and R methods to be used to obtain parameter estimates and predictions for the main effects of the outcome. See ?modelObj for details. |
data |
A data frame of the covariates and tx histories |
reward |
The response vector |
txName |
A character object. The column header of data that corresponds to the tx covariate |
regime |
A formula object or a list of formula objects. The covariates to be included in classification. If a list is provided, this specifies that there is an underlying subset structure – fSet must then be defined. |
response |
A numeric vector. The reward. Allows for naming convention followed in most DynTxRegime methods. |
fSet |
A function or NULL defining subset structure |
lambdas |
A numeric object or a numeric vector object giving the penalty tuning parameter. If more than 1 is provided, the finite set of values to be considered in the cross-validation algorithm |
cvFolds |
If cross-validation is to be used to select the tuning parameters, the number of folds. |
kernel |
A character object. must be one of {"linear", "poly", "radial"} |
kparam |
A numeric object of NULL. If kernel = linear, kparam is ignored. If kernel = poly, kparam is the degree of the polynomial If kernel = radial, kparam is the inverse bandwidth of the kernel. If a vector of bandwidth parameters is given, cross-validation will be used to select the parameter |
responseType |
A character indicating if response is continuous, binary or count data. |
verbose |
An integer or logical. If 0, no screen prints are generated. If 1, screen prints are generated with the exception of optimization results obtained in iterative algorithm. If 2, all screen prints are generated. |
an RWL object
Xin Zhou, Nicole Mayer-Hamblett, Umer Khan, and Michael R Kosorok (2017) Residual weighted learning for estimating individualized treatment rules. Journal of the American Statistical Association, 112, 169–187.
Other statistical methods:
bowl(),
earl(),
iqLearn,
optimalClass(),
optimalSeq(),
owl(),
qLearn()
Other weighted learning methods:
bowl(),
earl(),
owl()
Other single decision point methods:
earl(),
optimalClass(),
optimalSeq(),
owl(),
qLearn()
## Not run: # Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') fitRWL <- rwl(moPropen = moPropen, moMain = moMain, data = bmiData, reward = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, kernel = 'radial', kparam = 1.5) ##Available methods # Coefficients of the regression objects coef(fitRWL) # Description of method used to obtain object DTRstep(fitRWL) # Estimated value of the optimal treatment regime for training set estimator(fitRWL) # Value object returned by regression methods fitObject(fitRWL) # Summary of optimization routine optimObj(fitRWL) # Estimated optimal treatment for training data optTx(fitRWL) # Estimated optimal treatment for new data optTx(fitRWL, bmiData) # Value object returned by outcome regression method outcome(fitRWL) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(fitRWL) plot(fitRWL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitRWL) # Parameter estimates for decision function regimeCoef(fitRWL) # Show main results of method show(fitRWL) # Show summary results of method summary(fitRWL) ## End(Not run)## Not run: # Load and process data set data(bmiData) # define the negative 12 month change in BMI from baseline y12 <- -100*(bmiData[,6L] - bmiData[,4L])/bmiData[,4L] # propensity model moPropen <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'glm', solver.args = list('family'='binomial'), predict.method = 'predict.glm', predict.args = list(type='response')) # outcome model moMain <- buildModelObj(model = ~parentBMI+month4BMI, solver.method = 'lm') fitRWL <- rwl(moPropen = moPropen, moMain = moMain, data = bmiData, reward = y12, txName = 'A2', regime = ~ parentBMI + month4BMI, kernel = 'radial', kparam = 1.5) ##Available methods # Coefficients of the regression objects coef(fitRWL) # Description of method used to obtain object DTRstep(fitRWL) # Estimated value of the optimal treatment regime for training set estimator(fitRWL) # Value object returned by regression methods fitObject(fitRWL) # Summary of optimization routine optimObj(fitRWL) # Estimated optimal treatment for training data optTx(fitRWL) # Estimated optimal treatment for new data optTx(fitRWL, bmiData) # Value object returned by outcome regression method outcome(fitRWL) # Plots if defined by regression methods dev.new() par(mfrow = c(2,4)) plot(fitRWL) plot(fitRWL, suppress = TRUE) # Value object returned by propensity score regression method propen(fitRWL) # Parameter estimates for decision function regimeCoef(fitRWL) # Show main results of method show(fitRWL) # Show summary results of method summary(fitRWL) ## End(Not run)
RWL
Class RWL contains results for an RWL analysis.
responseTypecharacter indicating type of response
residualsvector of outcome residuals
betavector of regime parameters
analysisContains a Learning or LearningMulti object
analysis@txInfoFeasible tx information
analysis@propenPropensity regression analysis
analysis@outcomeOutcome regression analysis
analysis@cvInfoCross-validation analysis if single regime
analysis@optimOptimization analysis if single regime
analysis@optimResultlist of cross-validation and optimization results if multiple regimes. optimResult[[i]]@cvInfo and optimResult[[i]]@optim
analysis@optimalEstimated optimal Tx and value
analysis@CallUnevaluated Call
: Retrieve value object returned by outcome regression methods.
: Retrieve value object returned by propensity regression methods.
: Retrieve parameter estimates for all regression steps.
: Retrieve value object returned by regression methods.
: Generate plots for regression analyses.
: Retrieve cross-validation results.
: Retrieve value object returned by optimization method(s).
: Retrieve estimated parameters for optimal tx regime.
: Retrieve description of method used to create object.
: Retrieve the estimated value of the estimated optimal regime for the training data set.
: Retrieve/predict the estimated decision functions and/or optimal tx.
: Print main results of analysis.
: Show main results of analysis.
: Retrieve summary information.
Retrieve the standard deviation of the residuals for the first-stage contrasts regression in the interactive Q-Learning algorithm.
sd(x, na.rm=FALSE) ## S4 method for IQLearnFS_C sd(x, na.rm=FALSE)sd(x, na.rm=FALSE) ## S4 method for IQLearnFS_C sd(x, na.rm=FALSE)
x |
An object of class |
na.rm |
logical. Should missing values be removed? |
Returns a list of the primary results, including regression results, optimization results, estimated tx and value, etc.
summary(object, ...)summary(object, ...)
object |
Value object returned by a statistical method |
... |
Optional additional inputs |
Methods are defined for all statistical methods implemented in DynTxRegime.
The exact structure of the returned list will vary depending on the statistical method.