如何在R中实现第一列与其余列的皮尔逊相关性分析？

Let's walk through how to scale up your existing code to run Pearson correlation between the first column and every other column in your data frame, plus handle column removal based on the results.

Step 1: Batch Run Correlation Tests #

First, we'll automate running cor.test() for the first column against each of the remaining columns. Using lapply() makes this straightforward:

# Define your reference column (the first column of your data frame)
reference_col <- my_data[, 1]

# Run Pearson correlation for every other column
cor_results <- lapply(my_data[, -1], function(current_col) {
  cor.test(reference_col, current_col, method = "pearson")
})

Step 2: Tidy Up the Results #

Raw test outputs are a bit messy, so let's convert them into a readable summary table with key metrics:

# Turn results into a clean data frame
cor_summary <- do.call(rbind, lapply(names(cor_results), function(col_name) {
  test_result <- cor_results[[col_name]]
  data.frame(
    Compared_Column = col_name,
    Correlation_Coefficient = round(test_result$estimate, 4),
    P_Value = round(test_result$p.value, 4),
    stringsAsFactors = FALSE
  )
}))

# Print the summary to review your results
print(cor_summary)

Step 3: Remove Columns Based on Your Condition #

Your original code removes a column if the correlation coefficient is < 0.05. Quick note: Correlation coefficients range from -1 to 1, so this condition targets columns with extremely weak positive correlation. If you actually meant to remove columns with non-significant correlations (p-value > 0.05), I've included that alternative below.

Option A: Remove Columns with Correlation Coefficient < 0.05 #

# Identify columns that match your condition
cols_to_drop <- cor_summary$Compared_Column[cor_summary$Correlation_Coefficient < 0.05]

# Remove them from the data frame (only if there are columns to drop)
if (length(cols_to_drop) > 0) {
  my_data <- my_data[, !colnames(my_data) %in% cols_to_drop]
}

Option B: Remove Columns with Non-Significant P-Value (> 0.05) #

If your goal is to keep only columns with statistically significant correlations, use this instead:

cols_to_drop <- cor_summary$Compared_Column[cor_summary$P_Value > 0.05]

if (length(cols_to_drop) > 0) {
  my_data <- my_data[, !colnames(my_data) %in% cols_to_drop]
}

Bonus: Tidyverse-Friendly Approach #

If you prefer using dplyr and purrr, here's a more concise alternative:

library(dplyr)
library(purrr)

# Generate correlation summary
cor_summary <- my_data %>%
  select(-1) %>%
  map_dfr(function(col) {
    res <- cor.test(my_data[[1]], col, method = "pearson")
    tibble(
      Compared_Column = colnames(col),
      Correlation_Coefficient = round(res$estimate, 4),
      P_Value = round(res$p.value, 4)
    )
  })

# Remove columns based on your condition
cols_to_drop <- cor_summary$Compared_Column[cor_summary$Correlation_Coefficient < 0.05]
my_data <- my_data %>% select(-all_of(cols_to_drop))

内容的提问来源于stack exchange，提问作者Sindhu.C