R/simulation.R

In microclustr: Entity Resolution with Random Partition Priors for Microclustering

## function to simulate data ##

#' Generates a simulated dataset based on a true partition
#'
#' @param true_L Vector of size max cluster size with number of clusters of each size   
#' @param nfields Number of fields
#' @param ncat Vector with number of categories per field
#' @param true_beta Distortion probability for the fields
#' @return Simulated data set
#' @export
#' @examples
#' truePartition <- c(2,2,2,2)
#' numberFields <- 2
#' numberCategories <- rep(5,2)
#' trueBeta <- 0.01
#' SimData(truePartition, numberFields, numberCategories, trueBeta)
SimData <- function(true_L, nfields, ncat, true_beta){
  true_M <- length(true_L)
  true_K <- sum(true_L)
  N <- sum(c(1:true_M)*true_L)#100
  id <- rep(1:true_K,times=rep(1:true_M,times=true_L))
  ## generate true entities
  y <- matrix(data = sample.int(n = ncat,size = true_K*nfields,replace = TRUE),
             nrow = true_K,ncol = nfields)
  data<-matrix(NA,nrow = N,ncol = nfields)
  for (i in 1:N){
    for (l in 1:nfields){
      if(runif(1) < true_beta){
        data[i,l] <- sample.int(n = ncat[l],size = 1)
      }else{
        data[i,l] <- y[id[i],l]
      }
    }
  }
  x <- DataRemap(data)
  return(x)
}