R语言 使用plm包基于事件研究法(event-study re

使用plm包基于事件研究法(event-study regression)进行平行趋势判断，附加计算clustered standard errors，需要单独两行代码

后面使用fixest包中的feols()函数了，这个函数里面可以指定cluster的字段，备忘录一下plm的用法

两种方法的回归结果是一致的

library(modelsummary)
library(fixest)
library(did)
library(dplyr)
library(tidyverse)
library(plm)
library(stargazer)
library(lmtest)

# read data
df <- read.csv("E:\\panel_data.csv", header = 1)
# generate the treatment variable
df <- df %>% mutate(treatment = ifelse(df$nrty > 14, 0, 1))
# run event study regression
g <- plm( formula = ntl ~ i(year, treatment, ref = 14), data = df, 
          model = "within", effect = "twoways", index = c("id", "year"))
summary(g)
coefplot(g)

cluster_method <- coeftest(g, vcov = function(x) 
  plm::vcovHC( x, method = 'arellano', cluster = 'group', type = 'HC3') )
cluster_method

参考资料：https://stats.stackexchange.com/questions/10017/standard-error-clustering-in-r-either-manually-or-in-plm#:_{:text=library%20%28multiwayvcov%29%20library%20%28lmtest%29%20data%20%28petersen%29%20m1%20%3C-,%28m1%2C%20}%20firmid%20%2B%20year%29%20coeftest%20%28m1%2C%20vcov_both_formula%29