R/crossfit_U.R

In statdivlab/tinyvamp: Modeling complex measurement error in microbiome experiments

crossfit_U <- function(W,
                       full_model,
                       null_model,
                       parallelize = FALSE){
  
  n <- nrow(W)
  
  if(!parallelize){
    log_us <- numeric(n)
    
    for(i in 1:n){
      # print(i)
      train_ind <- rep(TRUE,n)
      train_ind[i] <- FALSE
      log_us[i] <- calculate_U(W, full_model, null_model,
                               training_indicator = train_ind)
    }
  } else{
    log_us <- parallel::mclapply(1:n,
                                 function(i){
                                   train_ind <- rep(TRUE,n)
                                   train_ind[i] <- FALSE
                                   return(calculate_U(W,
                                                      full_model,
                                                      null_model,
                                                      training_indicator = train_ind))
                                   
                                 })
    log_us <- unlist(log_us)
  }
  
  return(exp(logsum::sum_of_logs(log_us) - log(n)))
}