跟你要求的可能不太一样，但可试试。请先 ssc install mipolate，然后

谢谢黄老师，我学习一下~

You may need to think about in terms of a case that there are no values till the end of your sample period. For example, id 4 from 2004 to 2017. Taking the average of leading and ending non-missing values in a time series of an individual cannot be applied to this scenario. Hope this helpful. Dr. Huang may have a double check the correctness of this code.

谢谢您！
请问如果是想根据第2年、第3年的数据，填补第1年的缺失数据用什么命令比较好呢？
您列举的carryforward好像只能由前推后。

重新排序一下不就得了。