一、基准回归
1.数据处理
xtset code year
gen time=(year>=某一年份）&！missing(year)
gen treated=(code>某一数值）&！missing(code)
2.回归
xtreg  *** did time treated tenure hours *** *** *** *** id.Year.id.Year.,fe r
二、平行趋势检验
gen ***=*** if code>year
gen *** =*** if code <=year
bysort year:egen ***=mean(***)
duplicates drop year,force
scatter *** year ,c(1)||scatter *** year ，c（1）
三、安慰剂检验
• cap erase "simulations dta"
permute did beta = b did] se=_se[did] df = e(df_r), reps (500) seed (123)
saving ("simulations.dta"):reghdfe *** did, absorb( code year) vce(robust)
• use"simulations.dta", clear
• gent value = beta / se
• gen p value = 2 * ttail(df, abs(beta/se))
• ssc install dpplot 如果你没有安装的话要安装它
• dpplot beta, title ("Estimator", size(*0.8)) label(, format(½4.3f) labsize(small))
title("Density", size(*0.8)) ylabel(, nogrid format(44.3f) labsize(small)) note("'')
caption("'') graphregion(fcolor(white))

四、改变政策时间
• xtset code year
• gen time = (year >= 某一时间)&!missing (year)
• gen treated = (code >某一数值)&! missing(code)
• diff *** t(treated) p(time) cov( *** ***等控制变量) robust report bs reps(100) 100可以改变