如何用Proc Report或Data Step实现分组变量的多列计数表格？

完全可以用Proc Report或更简洁的Data Step工作流替代多次SQL分组再连接的繁琐操作，以下是两种可行方案的示例：

先构造示例数据集 #

先准备一份模拟业务数据，方便你直接运行测试：

data have;
  input Department $ Building_code $;
  datalines;
HR A
HR A
HR B
Finance B
Finance C
IT A
IT A
IT B
IT C
IT C
;
run;

方案一：直接用Proc Report实现 #

通过Proc Report的compute块，在分组过程中直接完成不同Building的条件计数，无需额外预处理：

proc report data=have nowindows headline headskip;
  columns Department Count_A Count_B Count_C;
  
  define Department / group '部门';
  define Count_A / computed 'Building A 数量';
  define Count_B / computed 'Building B 数量';
  define Count_C / computed 'Building C 数量';
  
  /* 分组前初始化计数变量 */
  compute before Department;
    count_a = 0;
    count_b = 0;
    count_c = 0;
  endcomp;
  
  /* 逐行判断并累加对应Building的计数 */
  compute;
    if Building_code = 'A' then count_a + 1;
    else if Building_code = 'B' then count_b + 1;
    else if Building_code = 'C' then count_c + 1;
  endcomp;
  
  /* 分组结束后输出汇总结果 */
  compute after Department;
    line @1 Department $10. count_a 5. count_b 5. count_c 5.;
    count_a = 0; count_b = 0; count_c = 0;
  endcomp;
  
  /* 跳过原始数据行，只保留分组汇总行 */
  break after Department / skip;
run;

方案二：Data Step分组计数 + Proc Transpose转置 #

如果你的Building_code取值较多且可能新增，这种方式更灵活，无需手动维护每个列的逻辑：

/* 第一步：按部门和Building分组统计数量 */
data count_temp;
  set have;
  by Department Building_code;
  if first.Building_code then count = 0;
  count + 1;
  if last.Building_code then output; /* 只输出每组最后一行的累计值 */
run;

/* 第二步：转置为宽表，自动生成各Building对应的计数列 */
proc transpose data=count_temp out=want prefix=Count_;
  by Department;
  id Building_code; /* 用Building_code的值作为列名前缀 */
  var count;
run;

/* 可选：用Proc Report格式化展示结果 */
proc report data=want nowindows headline headskip;
  columns Department Count_A Count_B Count_C;
  define Department / group '部门';
  define Count_A / display 'Building A 数量';
  define Count_B / display 'Building B 数量';
  define Count_C / display 'Building C 数量';
run;

这两种方案都比多次SQL分组再连接高效，新增Building_code时：

• 方案一只需新增对应的Count_X列定义和compute块中的判断逻辑；
• 方案二更省心，无需修改代码，转置会自动识别新的Building_code并生成对应列。

内容的提问来源于stack exchange，提问作者Kristián Ővári