如何在R语言中格式化含国家与年份的数据以进行回归分析？

Got it, let's walk through how to run linear regressions with one explanatory variable against multiple time-series dependent variables in R. Here's a practical, step-by-step guide tailored to your needs:

1. Data Preparation #

First, let's make sure your data is structured correctly. Since you're working with time series, you can either use a data.frame (with a time column) or keep variables as ts objects. Below is an example of simulating data that matches your description—replace this with your actual dataset:

# Load tidyverse for data manipulation (optional but helpful)
library(tidyverse)

# Simulate time-series data (replace with your real data)
set.seed(123) # For reproducible results
time_periods <- seq(as.Date("2010-01-01"), as.Date("2023-12-01"), by = "month")

# Explanatory variable
explanatory_index <- ts(
  rnorm(length(time_periods), mean = 100, sd = 10),
  start = c(2010, 1), frequency = 12
)

# Dependent variables (GDP, market_size, plus an extra example: sales)
GDP <- ts(
  50 + 2*explanatory_index + rnorm(length(time_periods), sd = 5),
  start = c(2010, 1), frequency = 12
)
market_size <- ts(
  100 + 1.5*explanatory_index + rnorm(length(time_periods), sd = 8),
  start = c(2010, 1), frequency = 12
)
sales <- ts(
  20 + 0.8*explanatory_index + rnorm(length(time_periods), sd = 3),
  start = c(2010, 1), frequency = 12
)

# Combine into a data frame (easier for regression workflows)
df <- data.frame(
  time = time_periods,
  explanatory_index = as.numeric(explanatory_index),
  GDP = as.numeric(GDP),
  market_size = as.numeric(market_size),
  sales = as.numeric(sales)
)

2. Option 1: Run Regressions One-by-One #

If you want to examine each model in detail (e.g., check residuals, coefficients), run individual regressions using lm():

# Regression for GDP
model_gdp <- lm(GDP ~ explanatory_index, data = df)
summary(model_gdp) # View full stats, coefficients, p-values

# Regression for market_size
model_market <- lm(market_size ~ explanatory_index, data = df)
summary(model_market)

# Regression for sales
model_sales <- lm(sales ~ explanatory_index, data = df)
summary(model_sales)

3. Option 2: Batch Process All Dependent Variables #

For efficiency (especially if you have many dependent variables), use purrr to run all regressions at once and compile results:

# Define the list of dependent variable names
dep_vars <- c("GDP", "market_size", "sales")

# Run all regressions in a loop
models <- map(
  dep_vars,
  ~lm(reformulate("explanatory_index", response = .x), data = df)
)

# Name the models for clarity
names(models) <- dep_vars

# View summaries for all models
map(models, summary)

# Extract key metrics into a clean data frame (coefficients, p-values, R²)
results_summary <- map_dfr(models, function(model) {
  stats <- summary(model)
  tibble(
    coefficient = stats$coefficients[2, 1],
    p_value = stats$coefficients[2, 4],
    r_squared = stats$r.squared
  )
}, .id = "dependent_variable")

print(results_summary)

4. Critical Time-Series Considerations #

Since your data is time series, don't forget to check for autocorrelation in residuals—a common issue that can bias standard errors:

library(forecast)
# Check residuals for the GDP model (repeat for others)
checkresiduals(model_gdp)

# If autocorrelation exists, use Newey-West standard errors to correct it
library(sandwich)
library(lmtest)
coeftest(model_gdp, vcov = NeweyWest(model_gdp))

This adjusts for autocorrelation and heteroscedasticity, giving you more reliable p-values.

Feel free to tweak the code to match your actual data structure (e.g., if you're working directly with ts objects instead of a data frame). If you need help interpreting specific outputs or troubleshooting issues, let me know!

内容的提问来源于stack exchange，提问作者John Doe