R/Adaptive.R

In rekkasa/SimulationEvaluationHte: What the Package Does (One Line, Title Case)

# fitAdaptiveApproach <- function(
#   trainingDataset,
#   validationDataset,
#   nonLinearSettings,
#   linearSettings,
#   constantSettings
# ) {
#
#   nonLinearFit <- rms::lrm(
#     outcome ~ treatment + rms::rcs(riskLinearPredictor, 3) + treatment * rms::rcs(riskLinearPredictor, 3),
#     data = trainingDataset
#   )
#   linearFit <- rms::lrm(
#     outcome ~ treatment + riskLinearPredictor + treatment * riskLinearPredictor,
#     data = trainingDataset
#   )
#   constantFit <- rms::lrm(
#     outcome ~ treatment + riskLinearPredictor,
#     data = trainingDataset
#   )
#   testResult      <- rms::lrtest(nonLinearFit, linearFit)
#   pValueNonLinear <- testResult$stats[["P"]]
#   testResult      <- rms::lrtest(linearFit, constantFit)
#   pValueConstant  <- testResult$stats[["P"]]
#
#   adaptiveMethod <- case_when(
#     pValueNonLinear < .05                          ~ "nonLinear",
#     pValueNonLinear >= .05 & pValueConstant < .05  ~ "linear",
#     pValueNonLinear >= .05 & pValueConstant >= .05 ~ "constant"
#   )
#
#   if (adaptiveMethod == "nonLinear") {
#     dataUntreated <- trainingDataset %>%
#       dplyr::filter(treatment == 0)
#     dataTreated <- trainingDataset %>%
#       dplyr::filter(treatment == 1)
#
#     smoothControl <- SmoothHte::fitRcsHte(
#       data     = dataUntreated,
#       settings = nonLinearSettings
#     )
#     smoothTreatment <- SmoothHte::fitRcsHte(
#       data     = dataTreated,
#       settings = nonLinearSettings
#     )
#
#     predictedBenefit <- SmoothHte::predictSmoothBenefit(
#       p               = plogis(validationDataset$riskLinearPredictor),
#       smoothControl   = smoothControl,
#       smoothTreatment = smoothTreatment
#     )
#   } else if (adaptiveMethod == "linear") {
#     modelBasedHte <- SmoothHte::fitModelBasedHte(
#       data     = trainingDataset,
#       settings = linearSettings
#     )
#
#     predictedBenefit <- SmoothHte::predictBenefitModelBasedHte(
#       p             = plogis(validationDataset$riskLinearPredictor),
#       modelBasedFit = modelBasedHte
#     )
#   } else if (adaptiveMethod == "constant") {
#     modelBasedHte <- SmoothHte::fitModelBasedHte(
#       data     = trainingDataset,
#       settings = constantSettings
#     )
#
#     predictedBenefit <- SmoothHte::predictBenefitModelBasedHte(
#       p             = plogis(validationDataset$riskLinearPredictor),
#       modelBasedFit = modelBasedHte
#     )
#   }
#
#   return(predictedBenefit)
# }