计算组内最大值对应记录

【问题】

Proc SQL Group By Question

Hello All,

       I hope you can help me on this. In my dataset I have siblings and their addresses. I am trying to group all siblings by address, so I know all the siblings who reside at an address. Now in this group I only want to pick the youngest sibling and get that information. I am trying to do this in EG, I have the dataset sorted by mailing address, but I would like to know is there a way where I can get the youngest sibling in the group. (something like select top 1)

Here is my dataset

Address              name             age               

111 street1          child1           5

111 street1          child2           10

In this group I only want to output info for

child1

别人的回答:

proc sql noprint;

  create table want as

    select *

      from have

        group by Address

          having age eq min(age);

【回答】

上面给出的 SQL 其实不正确,这个问题的难点是 min 函数只能计算出最小值,但不能取出“最小值对应的记录”,再加上分组和关联,就会让问题更显复杂。其实也可以用 keep/ top/row_number,或者用窗口函数等方法,当然,这些方法都比较麻烦。

这个问题结合 SPL 就简单多了,SPL 中的 top 函数可以取出最大值对应的记录,比如 salary.top(3;amount) 取出 amount 最小的三条记录,salary.top(-1;amount) 取出 amount 最大的那条记录。下面的这句代码可以解决你的所有问题:


A

1

=T1.group(Address).(~.top(1;age)).union()


简单解释一下:

T1: T1 表

.group(Address):按 Address 对 T1 分组

.(~.top(1;age)):继续计算,对每组数据,取出 age 最小的记录。

.union(): 将各组数据合并。

集算器是结构化计算开发工具,和 JAVA 或报表工具很容易集成,可以参考:【Java 如何调用 SPL 脚本