R/doOneEval.R

In benkeser/r2weight: Summarizing associations with multivariate outcomes

#' .doOneEval
#' 
#' One evaluation of the optimal weight combination and prediction.
#' @param validRows List of validation rows indices
#' @param Y The outcome
#' @param X The covariates
#' @param object The optWeight object 
#' @param seed The random seed to set
#' @param return.IC Should influence functions be returned
#' @param parallel Boolean to indicate running in parallel
#' @param n.cores Number of cores to use
#' 

.doOneEval <- function(validRows, Y, X, object, seed, return.IC, 
                       parallel, n.cores, ...){
    trainY <- Y[-validRows,,drop = FALSE]
    trainX <- X[-validRows,, drop = FALSE]
    validY <- data.matrix(Y[validRows, ,drop = FALSE])
    validX <- X[validRows, , drop = FALSE]
    # fit the optWeight routine
    fit <- optWeight(
        Y = trainY, X = trainX, SL.library = object$SL.library, 
        return.IC = return.IC, family = object$family, 
        CV.SuperLearner.V = object$CV.SuperLearner.V, 
        seed = seed, whichAlgorithm = object$whichAlgorithm, 
        return.CV.SuperLearner = FALSE, parallel = parallel, n.cores = n.cores
    )
    # make validation combined outcome
    Y.combn <- validY%*%matrix(fit$SL.weight)
    # make validation combined prediction
    psiHat <- predict(fit, newdata = validX)
    return(cbind(Y.combn, psiHat))
}