Stata实现多重中介代码 及Bootstrap中介效应

样本数据截图（企业数据）：

use data.dta, clear 

* 控制控制变量 ControlVariables

global ControlVariables SIZE ROA LEV ATTR TOP1 ACADEMIC MARKET i.IND  i.year 

* 链式中介命令

capture program drop mediation
program mediation, rclass
  gsem (med1 <- x $ControlVariables) (med2 <- x $ControlVariables) (y <- med2 med1 x $ControlVariables), vce(cluster stkcd) nocaps 
  return scalar cie1 = _b[med1:x]*_b[y:med1]
  return scalar cie2 = _b[med2:x]*_b[y:med2]
  return scalar direct_effect=_b[y:x]              
  return scalar total_effect=_b[y:x]+_b[med1:x]*_b[y:med1]+_b[med2:x]*_b[y:med2]
  end
set seed 12345

* 运行命令

mediation

*Bootstrap 中介效应

bootstrap r(direct_effect) r(cie1) r(cie2)  r(total_effect), reps(20) : mediation
estat boot, bc percentile