R语言多元线性回归中多CSV文件合并报错及批量处理需求

1. 修复多文件合并问题 #

Base R的merge()仅支持两两合并，要批量合并多个数据框，可通过Reduce()循环调用merge()实现。同时建议批量读取文件并统一列名，避免合并后列名冲突：

方法1：Base R实现 #

setwd('D:/MLR/NEW')

# 定义要读取的文件列表及对应变量名
files <- c("SM.csv", "TEMP.csv", "PPT.csv", "WS.csv", "ET.csv")
var_names <- c("SM", "Temp", "Ppt", "WS", "ET")

# 批量读取文件并重命名非Year列（避免列名重复）
data_list <- mapply(function(file, var) {
  df <- read.csv(file)
  colnames(df)[colnames(df) != "Year"] <- paste0(var, "_", colnames(df)[colnames(df) != "Year"])
  df
}, files, var_names, SIMPLIFY = FALSE)

# 按Year批量合并所有数据框
merged_data <- Reduce(function(x, y) merge(x, y, by = "Year", all = TRUE), data_list)

方法2：Tidyverse更简洁实现 #

library(tidyverse)

setwd('D:/MLR/NEW')

# 批量读取并标准化列名
data_list <- tibble(
  file = c("SM.csv", "TEMP.csv", "PPT.csv", "WS.csv", "ET.csv"),
  var = c("SM", "Temp", "Ppt", "WS", "ET")
) %>%
  mutate(data = map2(file, var, ~read.csv(.x) %>% rename_with(~paste0(.y, "_", .x), -Year))) %>%
  pull(data)

# 合并数据
merged_data <- reduce(data_list, inner_join, by = "Year")

2. 多网格点批量回归分析 #

假设每个文件包含多个网格点列（如ST1、ST2、ST3...），可通过分组或循环对每个网格点单独执行回归：

方法1：Tidyverse分组处理（推荐） #

library(broom)

# 转换为长格式，按网格点分组
long_data <- merged_data %>%
  pivot_longer(
    cols = -Year,
    names_to = c("Variable", "Grid"),
    names_sep = "_",
    values_to = "Value"
  ) %>%
  pivot_wider(names_from = Variable, values_from = Value)

# 批量运行回归并提取系数
reg_results <- long_data %>%
  group_by(Grid) %>%
  nest() %>%
  mutate(
    model = map(data, ~lm(SM ~ Temp + Ppt + WS + ET, data = .x)),
    tidied = map(model, tidy)
  ) %>%
  unnest(tidied) %>%
  select(Grid, term, estimate, std.error, statistic, p.value)

方法2：Base R循环处理 #

# 获取所有网格点名称
grid_points <- unique(sub(".*_", "", colnames(merged_data)[-1]))

# 初始化结果容器
results_list <- list()

for (grid in grid_points) {
  # 提取当前网格点的对应列
  sm_col <- paste0("SM_", grid)
  pred_cols <- paste0(c("Temp", "Ppt", "WS", "ET"), "_", grid)
  
  # 构建回归公式
  formula <- as.formula(paste(sm_col, "~", paste(pred_cols, collapse = "+")))
  
  # 运行回归并提取结果
  model <- lm(formula, data = merged_data)
  coef_df <- as.data.frame(summary(model)$coefficients)
  coef_df$term <- rownames(coef_df)
  coef_df$Grid <- grid
  
  results_list[[grid]] <- coef_df
}

# 合并所有结果
reg_results <- do.call(rbind, results_list)
colnames(reg_results)[1:4] <- c("estimate", "std.error", "statistic", "p.value")
reg_results <- reg_results[, c("Grid", "term", "estimate", "std.error", "statistic", "p.value")]

3. 原脚本的关键问题修正 #

• 合并逻辑错误：merge()不支持一次性传入多个数据框，必须通过Reduce()循环合并
• 公式引用错误：原脚本用"df$st1"这类外部对象引用构建公式，应直接使用合并后数据框的列名
• 大小写不匹配：原数据列名为ST1，脚本中用st1，需保持大小写一致

保存结果 #

将最终的回归系数结果导出到CSV：

write.csv(reg_results, "grid_regression_coefficients.csv", row.names = FALSE)

内容的提问来源于stack exchange，提问作者tom