R语言中两个numeric变量相减返回NA的问题求助

算力阿峰

2026-9-30

问题原因

核心问题是代码存在数据框名称不一致的笔误：你新增的ControlPermute列是存储在households数据框中的，但计算null_avg_control时，筛选条件误写为df$treatment=='Control'，df数据框内不存在ControlPermute列，取子集后得到的是空向量或全NA向量，求均值后自然返回NA。
你可以单独运行head(households$ControlPermute[df$treatment=='Control'])验证，输出结果会显示全为NA值。

解决方法

把计算null_avg_control那一行的df改为households即可，修正后的完整代码如下：

set.seed(10003)
Niter <- 5
null_diff <- vector(length=Niter)
for (i in 1:Niter){
    households$ControlPermute <- sample(households$voted[households$treatment=='Control'], size=180002, replace=T)
    
    null_avg_treated <- mean(households$ControlPermute[households$treatment=='Neighbors'])
    # 修正此处的数据框名称
    null_avg_control <- mean(households$ControlPermute[households$treatment=='Control'])
    null_diff[i] <- null_avg_treated - null_avg_control

    print(null_avg_treated)
    print(null_avg_control)
    print(null_diff[i])
}