R语言嵌套循环拟合多预测因子多结局简单线性回归报错求助

错误根因 #

• 公式拼接逻辑错误：你在paste中直接使用固定字符串"j ~"，R不会将此处的j替换为循环变量存储的结局变量名，而是会尝试调用环境中名为j的变量，不存在对应变量就会触发长度不匹配的报错。
• 未显式指定拟合数据来源：全局环境变量调用容易出现变量冲突，建议统一存入数据框后传入lm的data参数。

修正后基础实现代码 #

set.seed(42)
n <- 100
# 所有变量统一存入数据框，避免环境变量冲突
df <- data.frame(
  age = runif(n, min=45, max=85),
  sex = factor(sample(c('Male','Female'), n, replace=TRUE, prob=c(.6, .4))),
  smoke = factor(sample(c("Never", 'Former', 'Current'), n, replace=TRUE, prob=c(.25, .6, .15))),
  bmi = runif(n, min=16, max=45),
  outcome1 = rbinom(n = 100, size = 1, prob = 0.3),
  outcome2 = rbinom(n = 100, size = 1, prob = 0.13)
)

predictorlist <- c("age","sex", "bmi", "smoke")
outcomelist <- c("outcome1","outcome2")

for (i in predictorlist){ 
  for (j in outcomelist){ 
    # 正确拼接公式，调用循环变量存储的变量名
    model <- lm(formula = paste(j, "~", i), data = df) 
    # 打印模型标识方便对照结果
    cat("===== 拟合模型：", j, "~", i, "=====\n")
    print(summary(model))
  } 
}

Tidyverse风格可选实现 #

如果习惯tidyverse语法，可以用purrr包的交叉映射实现，后续批量提取回归结果也更方便：

library(purrr)

cross_df(list(predictor = predictorlist, outcome = outcomelist)) |> 
  pwalk(function(predictor, outcome){
    model <- lm(formula = paste(outcome, "~", predictor), data = df)
    cat("===== 拟合模型：", outcome, "~", predictor, "=====\n")
    print(summary(model))
  })

补充说明 #

你的结局变量为二分类0/1变量，当前用lm拟合的是线性概率模型，如果需要拟合逻辑回归，将lm替换为glm(formula = ..., data = df, family = binomial(link = "logit"))即可。

内容的提问来源于stack exchange，提问作者Sandro