面板数据，固定效应，回归时加不加xi为什么结果不同？

附件中面板数据，gvkey是公司代码，fyear是时间变量（年份），回归中需要加入 firm fixed effect 和 year fixed effect.

我做了如下两种操作：

1）
xtset gvkey fyear
xi: xtreg f.labrprod1 labrprod1 i.fyear, fe r cluster(gvkey)
2）
xtset gvkey fyear
xtreg f.labrprod1 labrprod1 i.fyear, fe r cluster(gvkey)


两种不应该表达同样的意思吗？为什么估计出来的截距项不一样？哪一种是正确的写法？

已阅读过help xi的文件，但是仍然不明白，求高人指点。非常感谢！

upload.dta (2.86 MB)

2019-1-18 17:28:44 上传

我自己想明白了，因为两种方法其实估计结果是相同的，只是用的baseline year不一样，所以截距项不一样，year dummy的系数不一样。btw,怎么删帖啊

The first command creates a series of dummy based on year. E.g. '_Ifyear_1972' treat observations in 1972 as 1 while the rest is valued in 0. Then you got group dummies for the regression and include year FE only. The second command uses the time series, i.e. one dummy for each year. And you use both year and firm FE.

Hope it's helpful to you.

réussite 发表于 2019-1-18 18:35
The first command creates a series of dummy based on year. E.g. '_Ifyear_1972' treat observations in ...

I think both commands included both year and firm FE. They are exactly the same.