如何在R原DataFrame中标记按年随机抽取的样本？

Hey there! Let's work through this problem—since you're handling large biological archival data, we need a clean, memory-efficient way to mark your selected rows directly in the original DataFrame, no more messy duplicates or Excel workarounds.

Core Idea #

Instead of extracting subsets and merging back (which causes duplicate issues), we'll add a marker column directly to your original dataset by grouping samples by year, then randomly selecting the required number of entries per group and tagging them.

Solution 1: Using dplyr (Clean & Readable) #

This is great for code clarity, and it’s efficient even for large datasets like yours.

First, let’s adapt your sample code to mark selected rows:

# Set seed for reproducible results (critical for research!)
set.seed(123)

# Your sample dataset
dat <- data.frame(
  X = sample(2000:2016, 50, replace=TRUE),
  Y = sample(c("yes", "no"), 50, replace = TRUE),
  Z = sample(c("french","german","english"), 50, replace=TRUE)
)

# Load dplyr (install first with install.packages("dplyr") if needed)
library(dplyr)

# Add "selection" column: mark 1 for randomly selected rows per year, 0 otherwise
dat_marked <- dat %>%
  group_by(X) %>%  # Group by year (your X column)
  mutate(
    selection = ifelse(
      row_number() %in% sample(n(), min(n(), 1)),  # 1 = number of samples per year; replace with 53 for your project
      1,
      0
    )
  ) %>%
  ungroup()

For Your Actual 150k-Row Dataset

Just replace the 1 in sample(n(), min(n(), 1)) with 53. The min(n(), 53) ensures you don’t get errors if a year has fewer than 53 samples (it will select all available entries for that year):

dat_marked_large <- dat %>%
  group_by(X) %>%
  mutate(
    selection = ifelse(
      row_number() %in% sample(n(), min(n(), 53)),
      1,
      0
    )
  ) %>%
  ungroup()

Solution 2: Base R (No Extra Packages) #

If you prefer sticking to base R, use the ave() function to handle grouping and tagging:

set.seed(123)

dat$selection <- ave(
  1:nrow(dat),  # Pass row numbers to group by year
  dat$X,        # Group by your year column
  FUN = function(x) {
    # Randomly select up to 53 rows per year
    selected_rows <- sample(x, min(length(x), 53))
    # Mark selected rows as 1, others as 0
    as.integer(x %in% selected_rows)
  }
)

How to Use This #

• Check selected rows: Simply filter the dataset to see which entries are marked:

  # View all selected samples
  selected_samples <- dat_marked %>% filter(selection == 1)
  
  # Verify a specific year's selections
  dat_marked %>% filter(X == 2000)
  

• Modify selected samples: If you need to edit fields for selected rows, do it directly in the original dataset (no merging required!):

  dat_marked <- dat_marked %>%
    mutate(Y = ifelse(selection == 1, "selected_yes", Y))
  

Key Benefits #

1. No duplicate rows: We’re modifying the original dataset in-place (well, creating a single copy) instead of merging subsets.
2. Memory-efficient: Works smoothly with your 150k-row dataset—no need to load huge Excel files.
3. Reproducible: Using set.seed() ensures your random selections stay consistent for future checks or collaborations.

内容的提问来源于stack exchange，提问作者C.Den