如何在R中针对不同educ值检验回归模型的dem弹性显著性？

Got it, let's break this down. Your elasticity formula for dem is:

Elasticity = β_log.dem + β_{log.dem_educ} * educ

You want to test if this elasticity equals 0 (i.e., is statistically significant) for educ values from 1 to 10. Your existing code works for educ=1 because the hypothesis simplifies to log.dem + log.dem_educ = 0—but for other values, we just need to adjust the linear constraint in linearHypothesis.

Step 1: Understand the General Hypothesis #

For any educ value k, the null hypothesis we want to test is:

β_log.dem + β_{log.dem_educ} * k = 0

We can translate this into a string that linearHypothesis understands by plugging k directly into the constraint.

Step 2: Implement the Tests (with Robust SEs) #

First, make sure you have the required packages loaded (car for linearHypothesis, sandwich for robust standard errors):

library(car)
library(sandwich)

Then, define your target educ values and run the tests in a loop (or with lapply for cleaner code):

Option 1: Loop through each educ value

# Define the range of educ values to test
educ_values <- 1:10

# Initialize a list to store test results
elasticity_test_results <- list()

for (k in educ_values) {
  # Build the hypothesis string dynamically
  hypothesis <- paste0("log.dem + ", k, "*log.dem_educ = 0")
  
  # Run the F-test with HC1 robust standard errors (matches your original code)
  test_out <- linearHypothesis(reg7, hypothesis, vcov = vcovHC(reg7, "HC1"))
  
  # Store the result with a name corresponding to the educ value
  elasticity_test_results[[as.character(k)]] <- test_out
}

Option 2: Use lapply for a more concise approach

educ_values <- 1:10

elasticity_test_results <- lapply(educ_values, function(k) {
  hypothesis <- paste0("log.dem + ", k, "*log.dem_educ = 0")
  linearHypothesis(reg7, hypothesis, vcov = vcovHC(reg7, "HC1"))
})

# Name the list elements for clarity
names(elasticity_test_results) <- educ_values

Step 3: Access and Analyze Results #

To view the test for a specific educ value (e.g., educ=5), just call:

elasticity_test_results[["5"]]

The key value to check is the Pr(>F) row—if this is below your significance threshold (usually 0.05), you can reject the null hypothesis and conclude the elasticity is statistically significant at that educ level.

Bonus: Organize Results into a Data Frame #

For easier comparison across all educ values, you can extract the key statistics into a tidy data frame:

results_summary <- do.call(rbind, lapply(educ_values, function(k) {
  hypothesis <- paste0("log.dem + ", k, "*log.dem_educ = 0")
  test_out <- linearHypothesis(reg7, hypothesis, vcov = vcovHC(reg7, "HC1"))
  
  data.frame(
    educ = k,
    F_statistic = round(test_out$F[2], 3),
    p_value = round(test_out$`Pr(>F)`[2], 4),
    statistically_significant = ifelse(test_out$`Pr(>F)`[2] < 0.05, "Yes", "No")
  )
}))

print(results_summary)

This will give you a clean table showing which educ values have a significant elasticity for dem.

内容的提问来源于stack exchange，提问作者Hassan A