Question

给出以下结构的表格：

subscriber_id，带

11,1

12,1

13,1

...

21,2

22.2

23.2

24,2

...

N1，N

N 2，N

N3，N

...

下，n

我想从每个组中获得n％大小订阅者的子组。对于10％，我应该得到组1的10％，组2的10％...组n的10％。

Answer 1

听起来你想要一个分层的样本。您可以先在每个组中枚举，然后选择＆＃34; n＆＃34;你想要的记录。以下是在SQL Server中如何执行此操作的示例：

select t.id, t.band
from (select t.*,
             row_number() over (order by band_seqnum) as seqnum
      from (select t.*,
                   row_number() over (partition by band order by rand(checksum()) as band_seqnum,
                   count(*) over () as cnt
            from t
           ) t
     ) t
where band_seqnum <= 0.10 * cnt;

Answer 2

试试这个

Select * from 
(
Select *, NTILE(n%) over(partition by id order by id) 'R' from t)t
where t.R<=(n%)

SQL服务器查询top和group by

2 个答案: