sql还真没有想到方法，排个序取前几条观测倒是可以… 还有univariate过程可以输出极端值，也可以输出最大的几个和最小的几个

univariate过程，可能还需要自己手工从输出窗口去提取，比较麻烦。排序倒是可以！如果要是能有个函数就更好了！非常感谢您！

SAS官方论坛上看到的：

data have;
input id $ score @@;
datalines;  /*临时弄几个分数测试*/
S12 -20 S12  20 S12  10
S12  10 S12  20 S12 -50
S13 -40 S13  60 S13 -100
S14  10 S14  20 S14  0
S14 -30 S14 -20
;

/* change as needed to bring back desired ranking  */
%let start= 1;
%let end = 3;

/*  create table containing distinct id and value  -- needed b/c more than one id can have the same score */
proc sql;
create table dst_id_score as
select distinct id, score from have
order by id, score desc;  /* keep this sort order for ranking */
* NOTE: Table WORK.DST_ID_SCORE created, with 12 rows and 2 columns. ;

/*  assign ranking by id by score */
data temp;
set dst_id_score ;
by id;
if first.id then rnk=0;
rnk + 1;
run;
* NOTE: The data set WORK.TEMP has 12 observations and 3 variables. ;

/*  go back to original file and bring back all data meeting desired ranking  */
proc sql;
create table final_top_ranked as
select
  a.id
, a.score
, b.rnk as rank
from have a inner join temp b on a.id = b.id and a.score = b.score
where rnk between &start and &end
order by id, rnk;
* NOTE: Table WORK.FINAL_TOP_RANKED created, with 11 rows and 3 columns. ;

这一段里只要调start和end，那么不止前三名，原则上可以查询任何范围内的数据。