如何将多回归分析结果按指定格式导出至Excel？

Great question! There are a few solid ways to get your regression table (with significance stars intact) into Excel while preserving the clean format you like. Here are the most straightforward methods:

1. Use the modelsummary Package (Simplest & Most Flexible) #

The modelsummary package is built exactly for this kind of task—it generates publication-ready tables and supports direct Excel export out of the box.

First, install the package if you haven't already:

install.packages("modelsummary")

Then run this code to get your desired table in Excel:

library(modelsummary)

# Your existing model code
data("mtcars")
m1 <- lm(hp ~ disp, data = mtcars)
m2 <- lm(hp ~ disp + wt, data = mtcars)

# Generate table and export directly to Excel
modelsummary(
  list(m1, m2),
  output = "regression_results.xlsx",
  stars = TRUE,  # Adds significance stars based on p-values
  statistic = "({std.error})",  # Shows standard errors in parentheses below coefficients
  gof_map = c("r.squared", "adj.r.squared", "nobs", "rmse")  # Include the exact stats you want
)

This creates an Excel file with a table that matches your desired format perfectly—coefficients with stars, standard errors below them, and all the goodness-of-fit metrics you need.

2. Use texreg to Export HTML, Then Convert to Excel #

If you prefer sticking with texreg, you can export the table to HTML first, which Excel can read directly (or convert it to a data frame in R first):

library(texreg)
library(openxlsx)  # For writing Excel files

# Your models
data("mtcars")
m1 <- lm(hp ~ disp, data = mtcars)
m2 <- lm(hp ~ disp + wt, data = mtcars)

# Export to HTML
htmlreg(list(m1, m2), file = "reg_table.html", include.ci = FALSE)

# Option 1: Open the HTML file directly in Excel (it will auto-convert to a formatted table)
# Option 2: Read the HTML table into R and export to Excel
library(rvest)
table_df <- read_html("reg_table.html") %>% html_table(fill = TRUE) %>% .[[1]]
write_xlsx(table_df, "regression_results.xlsx")

3. Manually Build a Data Frame (For Full Control) #

If you want complete control over every aspect of the table, you can extract components from your models and build a data frame manually before exporting:

library(texreg)
library(openxlsx)
library(dplyr)

data("mtcars")
m1 <- lm(hp ~ disp, data = mtcars)
m2 <- lm(hp ~ disp + wt, data = mtcars)

# Extract model components
ext1 <- extract(m1)
ext2 <- extract(m2)

# Helper function to add significance stars
add_stars <- function(coef, pval) {
  stars <- case_when(
    pval < 0.001 ~ "***",
    pval < 0.01 ~ "**",
    pval < 0.05 ~ "*",
    TRUE ~ ""
  )
  paste0(round(coef, 2), " ", stars)
}

# Build coefficient rows
coef_df <- data.frame(
  Term = c("(Intercept)", "disp", "wt"),
  Model1 = c(
    add_stars(ext1$coef[1], ext1$coef.pvalues[1]),
    add_stars(ext1$coef[2], ext1$coef.pvalues[2]),
    ""
  ),
  Model1_SE = c(paste0("(", round(ext1$se[1], 2), ")"), paste0("(", round(ext1$se[2], 2), ")"), ""),
  Model2 = c(
    add_stars(ext2$coef[1], ext2$coef.pvalues[1]),
    add_stars(ext2$coef[2], ext2$coef.pvalues[2]),
    add_stars(ext2$coef[3], ext2$coef.pvalues[3])
  ),
  Model2_SE = c(paste0("(", round(ext2$se[1], 2), ")"), paste0("(", round(ext2$se[2], 2), ")"), paste0("(", round(ext2$se[3], 2), ")"))
)

# Build goodness-of-fit rows
stats_df <- data.frame(
  Term = c("R^2", "Adj. R^2", "Num. obs.", "RMSE"),
  Model1 = c(round(ext1$rsquared, 2), round(ext1$adj.rsquared, 2), ext1$nobs, round(ext1$rmse, 2)),
  Model1_SE = rep("", 4),
  Model2 = c(round(ext2$rsquared, 2), round(ext2$adj.rsquared, 2), ext2$nobs, round(ext2$rmse, 2)),
  Model2_SE = rep("", 4)
)

# Combine and format the table
final_df <- rbind(coef_df, stats_df) %>%
  mutate(
    Model1 = ifelse(Model1_SE != "", paste0(Model1, "\n", Model1_SE), Model1),
    Model2 = ifelse(Model2_SE != "", paste0(Model2, "\n", Model2_SE), Model2)
  ) %>%
  select(Term, Model1, Model2)

# Add significance note at the bottom
note_df <- data.frame(
  Term = "*** p < 0.001, ** p < 0.01, * p < 0.05",
  Model1 = "",
  Model2 = ""
)
final_df <- rbind(final_df, note_df)

# Export to Excel
write_xlsx(final_df, "regression_results.xlsx")

In Excel, enable "Wrap Text" on the columns to see standard errors below coefficients, matching your original screenreg output.

内容的提问来源于stack exchange，提问作者msh855