917968079 发表于 2024-1-23 10:48 
2,3都只控制了行业和年份没有控制个体,第三个在absorb里面加入Stkcd
大神,感谢您的指导!
想跟您继续请教:
使用以下命令固定行业、年份和上市公司个体,以上市公司个体层面聚类稳健标准误,继续回归:
reghdfe tfp xzjj ROA LEV TBQ BOTH SIZE CASH1 BOARD SOE, absorb(Year ind Stkcd) vce(cluster Stkcd)
出现结果如下:
1.提醒 = FE nested within cluster; treated as redundant for DoF computation,损失太多自由度,是否有误?
2.回归结果继续不显著,P值=0.188,莫非我的模型有错,该怎么抢救?
3.咨询另外一位版友,提醒我尝试采用控制交互固定效应
尝试了 reghdfe tfp xzjj ROA LEV TBQ BOTH SIZE CASH1 BOARD SOE, absorb(Year ind Stkcd#Year) vce(cluster Stkcd)
直接报错:insufficient observations
| Robust
tpf | Coefficient std. err. t P>|t| [95% conf. interval]
-----------------+----------------------------------------------------------------
xzjj | .0700747 .0532049 1.32 0.188 -.0342451 .1743946
Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
Year | 8 1 7 |
ind | 20 1 19 |
Stkcd | 3149 3149 0 *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation