Create new objects from a for loop

[Zampano]

My function below works the way I want. I also want it to save the contents of filtered_xfile either to new objects, or to a list of objects (they are all different lengths so I'm not sure if that works).

I can't figure out how to save them with different names. For example fitlered_xfile1, filtered_xilfe2, filtered_xfile3. It prints them out just fine, but obviously overwrites when I save to an object and I just end up with the last xfile in the loop.

items <- c('HG13_2', 'GLAI18_2', 'HG15_2')

zscore_misfit <- function(data, xfile, item, sid) {
  xfile <- xfile %>%
    filter((xfile$ZSCORE > z_high | xfile$ZSCORE < z_low) & 
             xfile$`ITEM LABEL` == item)
  
  filtered_xfile <<- xfile
  
  data <<- data %>% 
    mutate("{item}" := replace(.data[[item]], .data[[sid]] %in% xfile$`PERSON LABEL`, 7774))
}

multiple_zscore <- function(xxx) {
  for (i in 1:(length(xxx))) {
    data <- zscore_misfit(data, xfile, xxx[i], "sidtp")
    print(items[i])
    print(filtered_xfile)
  }
}

test <- multiple_zscore(items)

[thomascf]

You're using the assign operator (<<-) to store object 'xfile' in the global environment as an object named 'filtered_xfile'. This process is repeated in every iteration of the for loop in multiple_zscore, each time overwriting the previous result. That's one of the hazards you'll need to watch out for when working with assign. Since your call to zscore_misfit happens within the function multiple_zscore, you could just do something like

multiple_zscore <- function(xxx){
  # Allocate an empty list to store results
  result_list <- list()

  for (i in seq_len(length(xxx))){
    data <- zscore_misfit(data, xfile, xxx[i], "sidtp")

    # Note: data and xfile are not provided as arguments in the function call.
    # Perhaps because they can be assumed to exist in the global environment
    # during runtime. However, I'd recommend adding these as arguments to the 
    # function definition of multiple_zscore.
    # e.g. multiple_zscore <- function(xxx, data, xfile, sid = "sidtp"){...}

    # Store result from zscore_misfit in result_list in ith entry.
    result_list[[i]] <- data
  }
  # Return list of results.
  return(result_list)
}

If you really do need to have each of these objects in the global environment (following your naming convention filtered_xfile1, filtered_xfile2, ...), you could do

multiple_zscore <- function(xxx){
  for (i in seq_len(length(xxx))){
    data <- zscore_misfit(data, xfile, xxx[i], "sidtp")
    new_name <- paste0("filtered_xfile",i)
    
    # Store object 'data' under the name 'new_name' in the global environment.
    assign(new_name, data)
  }
}

[elmstedt]

Complete aside, rather than seq_len(length(x)) you can simply use seq_along(x).

If you are going to enter elements into a list from a for() loop, the best way to do that is to pre-allocate your result list and seq_along() that, e.g.

res <- vector(mode = "list", length(xxx))
for (i in seq_along(res)) {
    # do stuff...
    res[[i]] <- ....
}

Better, though would be to forego the explicit loop and simply use lapply().

res <- lapply(xxx, zscore_misfit, data = data, xfile = xfile, sid = "sidtp")

All of this is ignoring the fact that it would probably just be better and easier to rewrite the original function to be properly vectorized itself.