按分组对DataFrame多列执行插值平滑滚动均值的高效R代码问询

Hey there! Let's work through this to get you the most efficient R code for your grouped column operations.

Since you need to apply a sequence of interpolation, smoothing, and rolling mean to multiple columns (debt_GDP, Top10) grouped by year and country_name_iso3, here are two optimized approaches—one using the tidyverse (great for readability) and another using data.table (ideal for large datasets):

First, define the core processing function #

We'll wrap your three steps into a reusable function so we can apply it cleanly to multiple columns:

library(zoo)
library(stats)

process_column <- function(x) {
  # Step 1: Spline-based NA interpolation
  interpolated <- na.interpolation(x, option = "spline")
  # Step 2: Smooth with spline regression
  smoothed <- fitted(smooth.spline(interpolated))
  # Step 3: 10-period rolling mean (fill missing edges with NA)
  rolled <- rollmean(smoothed, k = 10, fill = NA)
  return(rolled)
}

Note: If you need to adjust the smoothness of smooth.spline, add arguments like spar = 0.5 (tighter fit) or df = 5 (specified degrees of freedom) inside the function.

Option 1: Tidyverse (dplyr) Approach #

This is clean, readable, and works seamlessly with grouped data:

library(dplyr)

# Replace `your_df` with your actual DataFrame
processed_df <- your_df %>%
  group_by(year, country_name_iso3) %>%
  # Apply the function to target columns, name new columns clearly
  mutate(across(c(debt_GDP, Top10), process_column, .names = "{col}_final")) %>%
  ungroup()

Option 2: data.table Approach (Faster for Large Datasets) #

If you're working with a big dataset, data.table's grouped operations are more memory-efficient and faster:

library(data.table)

# Convert to data.table (skip if your data is already a data.table)
setDT(your_df)

processed_dt <- your_df[, 
                        # Create new columns by applying the function to each target column
                        c(paste0(c("debt_GDP", "Top10"), "_final")) := lapply(.SD, process_column),
                        # Group by your two variables
                        by = .(year, country_name_iso3),
                        # Specify which columns to process
                        .SDcols = c("debt_GDP", "Top10")
                        ]

Key Notes for Efficiency: #

• Both methods avoid manual column looping, which is notoriously slow in R.
• If some groups have fewer than 10 observations, the rolling mean will return all NAs for those groups. You can filter out small groups first with filter(n() >= 10) before grouping if needed.
• Make sure you've loaded the required packages (zoo, stats, plus either dplyr or data.table) before running the code.

内容的提问来源于stack exchange，提问作者msh855