R/testRF.R

In BAGofT: A Binary Regression Adaptive Goodness-of-Fit Test (BAGofT)

#' @export
#######################################################
# 'testModel' function for random forest
#######################################################
testRF <- function(formula, ntree = 500, mtry = NULL, maxnodes = NULL){
  # return a function of train data and test data
  testModel <- function(Train.data, Validation.data){
    if (is.null(mtry)){
      mtry <- max(floor( length( dim(stats::model.matrix(formula ,Train.data))[2]-1)/3),1)
    }

    # obtain the response name
    Rsp <- as.character(formula)[2]

    RspDat <- Train.data[,Rsp]

    Train.data[,Rsp] <- as.factor(Train.data[,Rsp])
    Validation.data[,Rsp] <- as.factor(Validation.data[,Rsp])

    resRf <- randomForest :: randomForest(formula, data = Train.data, ntree = ntree,  maxnodes = maxnodes, mtry = mtry, importance=FALSE)
    # obtain random forest prediction on the training set
    predT <-  stats :: predict(resRf, newdata = Train.data, type =  "prob")[,2]
    # obtain random forest prediction on the test set
    predE <-  stats :: predict(resRf, newdata = Validation.data, type =  "prob")[,2]

    # calculate the Pearson residual
    # res <-   (RspDat -  predT)/sqrt(predT * (1 - predT ))
    res <-   (RspDat -  predT)


    return(list(predT = predT, predE = predE, res = res, Rsp = Rsp))
  }
  return(testModel)
}