optimalClass function

Classification Perspective

optimalClass(
  ...,
  moPropen,
  moMain,
  moCont,
  moClass,
  data,
  response,
  txName,
  iter = 0L,
  fSet = NULL,
  verbose = TRUE
)

Arguments

• ...: 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.

Returns

an object of class OptimalClass

Examples

# 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

References

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.

See Also

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()