你的时间数据好象有问提。不同的时间段可用proc format定义。如不考虑时间，可用下面的 proc sql。
data test;
 input date DTime TNumber;
datalines;
20070104 93000 1
20070104 93004 2
20070104 93010 1
20070104 93014 2
20070104 93020 2
20070104 93024 2
20070104 93030 1
20070104 93034 1
20070104 93040 0
20070104 93044 0
20070105 93050 0
20070105 93054 2
20070105 93100 2
20070105 93104 1
20070105 93110 1
20070105 93114 0
20070105 93120 1
20070105 93124 0
20070105 93130 2
20070105 93134 1
20070105 93140 2
20070105 93144 1
20070105 93150 2
20070105 93154 1
20070105 93200 0
;
run;
proc sql;
   select distinct date, Tnumber, count(tnumber) as count
   from test
   group by date, tnumber;
quit;

output
   date   TNumber     count
-----------------------------
20070104         0         2
20070104         1         4
20070104         2         4
20070105         0         4
20070105         1         6
20070105         2         5

感谢yongyitian的回复，时间93010表示9点30分10秒，所以应该是没问题的。

实际数据中9：30～9：31之间出现0的次数应该是3次、出现2的次数应该是5次，但是按照你的程序运行的结果出现0的次数只有2次、出现2的次数只有4次，不知道哪位高手能不能再改改？

谢谢大家。

[此贴子已经被作者于2009-2-28 12:57:26编辑过]

Is this what you want,

proc sql;
  select distinct date,
         case when dtime >=93000 and dtime<93100 then "p1 9:30:00 - 9:30:59"
              when dtime >=93100 and dtime<93200 then "p2 9:31:00 - 9:31:59"
              else "p3 others add here"
              end as Time_period,
         Tnumber,
         count(Tnumber) as count
 from test
 group by date, Time_period, Tnumber;
quit;

output
     date  Time_period            TNumber     count
---------------------------------------------------
20070104  p1 9:30:00 - 9:30:59         0         2
20070104  p1 9:30:00 - 9:30:59         1         4
20070104  p1 9:30:00 - 9:30:59         2         4
20070105  p1 9:30:00 - 9:30:59         0         1
20070105  p1 9:30:00 - 9:30:59         2         1
20070105  p2 9:31:00 - 9:31:59         0         2
20070105  p2 9:31:00 - 9:31:59         1         6
20070105  p2 9:31:00 - 9:31:59         2         4
20070105  p3 others add here           0         1