不知道这样是不是楼主想要的结果：

data have;
input Cat $ Val;
datalines;
A 1
A 2
C 88
B 3
C 4
C 5
C 6
;
run;

proc sql;
  create table wanted(drop=tmp) as
   select *,monotonic() as tmp
    from have
        group by cat
    having count(*)=1 or tmp>min(tmp)
    order by tmp;
quit;

好，谢谢

多个排序
proc sort data=have;
by cat;
run;

data want;
set have;
by cat;
if first.cat and ^last.cat then delete;
run;

看需求的话，数据的顺序应该是不能动的：
data want;
set have;
by cat notsorted;
if first.cat and ^last.cat then delete;
run;

对，就是这样，多谢pobel神

我只想到了得用个辅助变量，但是不知道这一步
create table wanted(drop=tmp) as

我还想着必须要再加一步删变量呢
还是sql用得不熟啊