如何在R中高效对数据框变量执行对数变换并新增对应列？

Great question! Manually creating each log-transformed column like your example works, but it’s repetitive—especially if you have more variables to process. Here are several efficient, scalable approaches to achieve exactly what you need:

1. Tidyverse (dplyr) Approach (Most Readable for Tidy Data Users) #

If you’re already using the tidyverse, the across() function (replacing the older mutate_at()) makes this task clean and intuitive. It lets you apply a function to multiple columns and automatically generate new column names using a simple template.

library(tidyverse)
data("mtcars")

# Apply log transform to specific columns and add as new "log{column}" columns
mtcars_transformed <- mtcars %>%
  mutate(
    across(
      .cols = c(disp, hp, wt, qsec),  # Pick the columns to transform
      .fns = ~log(.),                 # The log transformation function
      .names = "log{.col}"            # Naming rule for new columns
    )
  )

# Check the result
head(mtcars_transformed)

If you want to apply the log transform to all numeric columns instead of a specific list, just swap the .cols argument with where(is.numeric):

mtcars_transformed <- mtcars %>%
  mutate(across(where(is.numeric), ~log(.), .names = "log{.col}"))

2. data.table Approach (Fastest for Large Datasets) #

For large datasets, data.table is unbeatable for speed and memory efficiency. Its in-place modification operator (:=) lets you add new columns without copying the entire dataset—critical for big data workflows.

library(data.table)
data("mtcars")
dt_mtcars <- as.data.table(mtcars)

# Define columns to transform
cols_to_transform <- c("disp", "hp", "wt", "qsec")

# Add log-transformed columns in one line
dt_mtcars[, paste0("log", cols_to_transform) := lapply(.SD, log), .SDcols = cols_to_transform]

# Check the result
head(dt_mtcars)

The .SDcols argument tells data.table which columns to process, and paste0("log", cols_to_transform) dynamically generates your desired new column names.

3. Base R Approach (No External Packages Needed) #

If you prefer to avoid loading extra libraries, base R has simple ways to eliminate repetitive code too.

Using lapply() #

data("mtcars")
cols_to_transform <- c("disp", "hp", "wt", "qsec")

# Generate log-transformed columns as a list
log_transformed_cols <- lapply(mtcars[cols_to_transform], log)
# Rename list elements to match your naming rule
names(log_transformed_cols) <- paste0("log", cols_to_transform)
# Bind new columns to the original data frame
mtcars_transformed <- cbind(mtcars, log_transformed_cols)

head(mtcars_transformed)

Using a for Loop #

This is straightforward and easy to follow, even for those new to R:

data("mtcars")
cols_to_transform <- c("disp", "hp", "wt", "qsec")

# Loop through each column and add its log-transformed version
for(col in cols_to_transform) {
  mtcars[[paste0("log", col)]] <- log(mtcars[[col]])
}

head(mtcars)

Which One Should You Choose? #

• Go with the tidyverse approach if you prioritize readability and already work with tidy data tools.
• Use data.table if you’re handling large datasets and need maximum performance.
• Stick to the base R approach if you want to avoid dependencies or work in a minimal environment.

内容的提问来源于stack exchange，提问作者aelhak