R Sandwich 包是否没有生成预期的集群稳健标准错误?

Is R Sandwich package not generating the expected clustered robust standard errors?

加载数据

utils::data("InstInnovation", package = "sandwich")

df <- InstInnovation

创建组变量组合 'company' 和 'year'

df[['cluster_var']] <- factor(paste0(df$company,"-",df$year))

线性回归模型

model <- lm(sales ~ competition + log(capital/employment) + year, data = df)

为什么这样:

lmtest::coeftest(model, vcov = vcovCL(model, type="HC3", cluster=~company+year))

产生与此不同的标准错误?

lmtest::coeftest(model, vcov = vcovCL(model, type="HC3", cluster=~cluster_var))

cluster=~company+yearcluster=~cluster_var 不应该等价吗?

此外,我找不到一个地方(例如Github)报告关于R sandwich包的问题,​​我找到了这个但只是一个只读镜像:https://github.com/cran/sandwich

非常感谢您。

cluster=~company+year 确实有些不同:'multiway clustering'。我在这里找到了解释:

http://fmwww.bc.edu/repec/bost10/BOS10.baum.pdf

https://francish.netlify.app/post/note-on-robust-standard-errors/