如何以Tidy风格为数据框每行多列应用带独立参数的函数？

Hey there! Let's walk through how to solve this using tidyverse tools—no clunky for loops required. The goal is to apply a custom function that takes separate column values from each row as arguments, then populate the p_value1 and p_value2 columns with the results.

Step 1: Define Your Custom Function #

First, let's formalize the function that will calculate the two p-values. I'll use common statistical tests as examples, but you can swap this out for your actual logic:

calculate_p_values <- function(success, trial, prog) {
  # Example 1: Binomial test p-value
  p1 <- binom.test(x = success, n = trial, p = prog)$p.value
  # Example 2: Poisson test p-value (alternative metric)
  p2 <- poisson.test(x = success, lambda = trial * prog)$p.value
  # Return both values as a named list
  list(p_value1 = p1, p_value2 = p2)
}

Step 2: Apply the Function Row-Wise (Two Tidy Methods) #

Method 1: Using dplyr::rowwise() (Most Intuitive) #

This method is great for readability—we explicitly tell R to process the data row-by-row, then call our function with the relevant columns:

library(dplyr)

# Process the data
test_data_processed <- test_data %>%
  rowwise() %>% # Enable row-wise operations
  mutate(
    # Call our function with the three columns as arguments
    p_results = list(calculate_p_values(successes, trials, prognosis)),
    # Extract the two p-values into separate columns
    p_value1 = p_results$p_value1,
    p_value2 = p_results$p_value2
  ) %>%
  select(-p_results) %>% # Remove the temporary list column
  ungroup() # Revert back to standard data frame behavior

Method 2: Using purrr::pmap() (Functional Programming Style) #

If you prefer a more functional approach, pmap from the purrr package is perfect for mapping multiple arguments to a function across rows:

library(dplyr)
library(purrr)

test_data_processed <- test_data %>%
  mutate(
    # Map the three columns to our function's arguments
    p_results = pmap(
      .l = list(success = successes, trial = trials, prog = prognosis),
      .f = calculate_p_values
    )
  ) %>%
  unnest_wider(p_results) # Expand the list column into separate columns

Key Notes #

• Argument Matching: Make sure the order (or names) of the columns passed to the function match the function's parameter names. Using named lists in pmap (like success = successes) avoids mix-ups if your function has many parameters.
• Ungrouping: After using rowwise(), don't forget to ungroup() if you plan on doing further column-wise operations—otherwise, R will keep treating the data as row-grouped.
• Custom Logic: Replace the test calculations in calculate_p_values with your actual function logic—this structure works for any function that takes multiple scalar inputs and returns multiple outputs.

Example Output Preview #

After running either method, your test_data_processed will have populated p_value1 and p_value2 columns:

# A tibble: 10 × 5
   successes trials prognosis p_value1 p_value2
       <dbl>  <dbl>     <dbl>    <dbl>    <dbl>
 1         0     50    0.0005  1.00e 0  1.00e 0
 2         3   1788    0.001   9.77e-1  7.77e-1
 3         6   1876    0.0025  2.60e-1  1.98e-1
# … with 7 more rows

内容的提问来源于stack exchange，提问作者Richi W