如何用R循环重复执行抽样与调查分析代码30次？

Hey Paola, great question—manually writing that code 30 times would be such a tedious hassle. Let’s use R’s built-in tools to automate this process smoothly. Since your code relies on the survey package, here are two reliable, easy-to-implement approaches:

Approach 1: Using a for Loop (Intuitive for Beginners) #

This method is straightforward and easy to follow, perfect if you prefer explicit, step-by-step logic. We’ll loop through 30 iterations, run your sampling/analysis each time, and store all results for later review.

Full Code: #

# Load the required package (install first if you haven't: install.packages("survey"))
library(survey)

# Initialize empty lists to store our 30 sets of results
total_estimates <- vector("list", 30)
mean_estimates <- vector("list", 30)

# Define your population size (used in the finite population correction)
population_size <- 6399875

# Run the loop 30 times
for (i in 1:30) {
  # Step 1: Draw a 10% SRS sample from your data
  srs_indices <- sample(1:nrow(MyData_points), size = 0.10 * nrow(MyData_points))
  sampled_data <- MyData_points[srs_indices, ]
  
  # Step 2: Set up FPC (note: we use nrow(sampled_data) to match the actual sample size each time)
  fpc_values <- rep(population_size, nrow(sampled_data))
  
  # Step 3: Create the survey design object
  srs_design <- svydesign(id = ~1, strata = NULL, data = sampled_data, fpc = fpc_values)
  
  # Step 4: Calculate total and mean, then store the results
  total_estimates[[i]] <- svytotal(~V4, design = srs_design)
  mean_estimates[[i]] <- svymean(~V4, design = srs_design)
  
  # Optional: Print progress to track how many iterations are done
  cat("Finished iteration", i, "\n")
}

# Convert results to data frames for easier analysis (e.g., calculating averages of estimates)
total_results_df <- do.call(rbind, lapply(total_estimates, as.data.frame))
mean_results_df <- do.call(rbind, lapply(mean_estimates, as.data.frame))

Approach 2: Using purrr (Cleaner for Tidyverse Users) #

If you’re comfortable with the tidyverse ecosystem, using purrr::map lets you write more concise, function-focused code without explicit loops.

Full Code: #

library(survey)
library(purrr)
library(dplyr)

# Define a function that runs one complete iteration of your sampling/analysis
run_single_srs <- function() {
  # Draw sample
  srs_indices <- sample(1:nrow(MyData_points), size = 0.10 * nrow(MyData_points))
  sampled_data <- MyData_points[srs_indices, ]
  
  # Set up FPC and survey design
  fpc_values <- rep(6399875, nrow(sampled_data))
  srs_design <- svydesign(id = ~1, strata = NULL, data = sampled_data, fpc = fpc_values)
  
  # Return both total and mean estimates in a list
  list(
    total_estimate = svytotal(~V4, design = srs_design),
    mean_estimate = svymean(~V4, design = srs_design)
  )
}

# Run the function 30 times
all_srs_results <- map(1:30, ~run_single_srs())

# Extract and convert results to tidy data frames
total_results_df <- all_srs_results %>% 
  map("total_estimate") %>% 
  map_dfr(as.data.frame)

mean_results_df <- all_srs_results %>% 
  map("mean_estimate") %>% 
  map_dfr(as.data.frame)

Key Tips: #

• FPC Adjustment: Your original code uses rep(6399875, 639987)—I updated both approaches to use nrow(sampled_data) instead, which ensures the FPC vector always matches your actual sample size (critical if your data’s row count isn’t perfectly divisible by 10).
• Result Analysis: Once you have the results in data frames, you can easily calculate summary stats (like the average of all 30 total estimates, or their standard deviation) to assess how much your estimates vary across samples.
• Package Checks: Make sure you have all required packages installed before running the code—use install.packages(c("survey", "purrr", "dplyr")) if needed.

内容的提问来源于stack exchange，提问作者Paola