如何为data.table指定列的每个单元格乘以符合特定条件的随机因子

Let's break down how to achieve your goal of multiplying every cell in specified columns of a data.table with a unique random factor (between 0.5-1.5, excluding 1), even when the number of rows is unknown.

Step 1: Create a Robust Factor Generator #

First, we need a way to generate valid factors efficiently. Instead of throwing an error if a factor equals 1 (which is statistically extremely unlikely, but possible in edge cases), we'll regenerate any invalid values:

generate_factor_matrix <- function(n_row, n_col) {
  # Generate initial batch of random factors
  factor_mat <- matrix(runif(n_row * n_col, min = 0.5, max = 1.5), 
                       nrow = n_row, ncol = n_col)
  
  # Replace any accidental 1s (edge case handling)
  while(any(factor_mat == 1)) {
    invalid_idx <- which(factor_mat == 1)
    factor_mat[invalid_idx] <- runif(length(invalid_idx), 0.5, 1.5)
  }
  
  return(factor_mat)
}

Step 2: Apply the Factors to Your data.table #

Using data.table's efficient .SD (Subset of Data) syntax, we can directly multiply the target columns with our factor matrix:

library(data.table)

# Your original data.table
testDF <- data.table(col1 = c(1,1,1), col2 = c(2,2,2), col3 = c(3,3,3))
# Columns to modify
SC <- c("col1", "col3")

# Get dimensions of the target subset
subset_dims <- dim(testDF[, ..SC])
n_rows <- subset_dims[1]
n_cols <- subset_dims[2]

# Generate the factor matrix matching the subset dimensions
factor_mat <- generate_factor_matrix(n_rows, n_cols)

# Perform element-wise multiplication and update the original data.table
testDF[, (SC) := .SD * factor_mat, .SDcols = SC]

Why This Works #

• Efficiency: Batch-generating random factors is far faster than generating them one-by-one, especially with large datasets.
• data.table Optimization: Using .SD and .SDcols leverages data.table's optimized column-wise operations, avoiding slow loops.
• Robustness: The loop ensures no factor equals 1, covering even the rarest edge cases.
• Flexibility: This works regardless of the number of rows (since we dynamically fetch n_rows from the subset).

Alternative: Create a Factor data.table #

If you prefer working with a data.table instead of a matrix, you can convert the factor matrix and align column names:

factor_dt <- as.data.table(factor_mat)
setnames(factor_dt, SC)
testDF[, (SC) := .SD * factor_dt, .SDcols = SC]

This achieves the exact same result but might feel more intuitive if you're used to data.table structures.

内容的提问来源于stack exchange，提问作者Nneka