不知道你的unit of analysis 是社么

如果是firn的话，
那是3-level random-effect modeling

level-1 firm
level-2 industry
level-3 year

i will use xtmixed to run this three-level model if DV is a continuous variable, or xtmelogit if DV is a binary one.

xtmixed y x1 x2 ||year: || industry: || firm:



if your unit of analysis is industry,

xtset industry year
xtreg y x1 x2, re


whether or not to include pannel variable(s) as your dummies depends on your research Q.


if research interest focuses on policy effects, i would not control for either time (year) or panel variable (industry). but I will explain variance componets in the multilevel model. it could tell us how much variance in each level.


i hardly see authors controled pannel variables as dummies in their panel analysis or multilevel model analysis, at least in my field, unless there is a spesiifc interest on effects of panel variable.

如果x是根据年份来划分0和1的，再放年份的dummy当然会被omit掉。

那是不用再放年份的dummy吗？我看很多文章虽然x也是按照年份划分的，都仍旧放年度dummy的，解释是控制控制不同年度宏观经济政策

这个问题值得好好讨论讨论，求大神出来解释一番

楼主请问您最后怎么处理的，我也出现了一样的问题