I have a dataset, which is a list of pairwise comparison stimuli I want to randomly and evenly allocate to subsetted groups, but with some rules.
The comparison data frame is as below:

structure(list(comparison = c(1, 2, 3, 4, 5, 6), `speaker 1` = c("a", 
"a", "a", "a", "a", "a"), `condition 1` = c(1, 1, 1, 1, 1, 1), 
    `voice 1` = c("a1", "a1", "a1", "a1", "a1", "a1"), `speaker 2` = c("b", 
    "c", "d", "e", "b", "c"), `condition 2` = c(1, 1, 1, 1, 2, 
    2), `voice 2` = c("b1", "c1", "d1", "e1", "b2", "c2")), row.names = c(NA, 
-6L), class = c("tbl_df", "tbl", "data.frame"))

there are 360 comparisons that I want to subset into 6 equal groups of exactly 60 comparisons.
I want to make sure therefore that firstly all unique comparisons are allocated to a subset; secondly that all subsets contain exactly 60 unique comparisons; and thirdly that all subsets contain exactly 4 comparisons for each voice, whether in ‘voice 1’ column or ‘voice 2’ column.

I have been troubleshooting some code for a few days, and have now been running the following for several hours with no errors, but it seems like it’s too computationally demanding and I’m wondering if there’s a workaround:

# Function to split the data into equal subsets with exactly 60 comparisons in each subset
split_data <- function(stimuli_manipulation, num_subsets) {
  # Shuffle the data
  stimuli_manipulation <- stimuli_manipulation[sample(nrow(stimuli_manipulation)), ]
  
  # Initialize an empty list to store subsets
  subsets <- vector("list", num_subsets)
  
  # Initialize counters for each voice
  voice_counts <- table(unlist(stimuli_manipulation[, c("voice 1", "voice 2")]))
  
  # Initialize an index for subsets
  subset_index <- 1
  
  # Initialize a counter for the total number of comparisons across subsets
  total_comparisons <- 0
  
  # Initialize an index for the current comparison
  i <- 1
  
  # Loop until all subsets have 60 comparisons
  while (total_comparisons < 60 * num_subsets) {
    # Check if the total number of comparisons across subsets reaches 60 * num_subsets
    if (total_comparisons >= 60 * num_subsets) {
      break
    }
    
    # Get the current comparison
    comparison <- stimuli_manipulation[i, ]
    
    # Check if there are no missing values in the comparison
    if (!is.na(comparison$`voice 1`) && !is.na(comparison$`voice 2`)) {
      # Check if adding the comparison maintains balance for each voice
      if ((voice_counts[comparison$`voice 1`] + voice_counts[comparison$`voice 2`]) < 4) {
        # Add the comparison to the subset
        subsets[[subset_index]] <- rbind(subsets[[subset_index]], comparison)
        
        # Update the counter for each voice
        voice_counts[comparison$`voice 1`] <- voice_counts[comparison$`voice 1`] + 1
        voice_counts[comparison$`voice 2`] <- voice_counts[comparison$`voice 2`] + 1
        
        # Increment the total number of comparisons
        total_comparisons <- total_comparisons + 1
      }
    }
    
    # Move to the next comparison
    i <- i %% nrow(stimuli_manipulation) + 1
    
    # Move to the next subset
    subset_index <- subset_index %% num_subsets + 1
  }
  
  # Save the subsets as data frames in the workspace and as CSV files
  for (i in 1:num_subsets) {
    assign(paste0("subset_", i), subsets[[i]], envir = .GlobalEnv)
    write.csv(subsets[[i]], paste0("subset_", i, ".csv"), row.names = FALSE)
  }
  
  # Return the list of subsets
  return(subsets)
}

# Call the function to split the data into subsets
subsets <- split_data(stimuli_manipulation, 6)

I also think I need to change it so that the ‘voice counts’ sums across voice 1 and 2, so when updating the count it’s accurate, but I’m figuring this out on my own and don’t have much experience!

Thanks in advance