ds.medianDP.R

source("R/utils.R")

#' @title Differentially private median
#'
#' @param input_data the input vector
#' @param epsilon privacy budget
#' @param lower_bound lower bound for input values
#' @param upper_bound upper bound for input values
#' @param datasources a list of \code{\link{DSConnection-class}} 
#' objects obtained after login. If the \code{datasources} argument is not specified
#' the default set of connections will be used: see \code{\link{datashield.connections_default}}.
#'
#' @return \code{ds.medianDP} returns a differentially private median
#' @export

ds.medianDP <- function(input_data, epsilon, lower_bound, upper_bound, type="split", datasources=NULL) {

  if (is.null(datasources)) {
    datasources <- DSI::datashield.connections_find()
  }
  if (!type %in% c("both", "split", "combine")) {
    stop("Type must be one of 'both', 'split' or 'combine'")
  }

  median.data <- callAggregationMethod(datasources, paste0("medianDP(", input_data, ", ", epsilon, ", ", lower_bound, ", ", upper_bound, ")"))

  Nstudies <- length(datasources)
  median.mat <- matrix(as.numeric(unlist(median.data)),nrow=Nstudies,byrow=TRUE)

  median.split <- median.mat
  dimnames(median.split) <- c(list(names(median.data), names(median.data[[1]])))

  median.combine <- t(matrix(median.mat[1,]))
  median.combine[1,1] <- ((t(matrix(median.mat[,2]))%*%median.mat[,1])/sum(median.mat[,2]))
  median.combine[1,2] <- sum(median.mat[,2])
  dimnames(median.combine) <- c(list("studiesCombined"),list(names(median.data[[1]])))

  if (type=="combine") return(list(Global.Median=median.combine,Nstudies=Nstudies))
  if (type=="split") return(list(Median.by.Study=median.split,Nstudies=Nstudies))
  if (type=="both") return(list(Median.by.Study=median.split,Global.Median=median.combine,Nstudies=Nstudies))
}