Question

我有一个如下所示的数据透视查询，得到的行count和AVG很好，但没有标准偏差STD。

如何修改下面的SQL以获取STD？

SELECT mid                             as mID,
   round((x.qty_sum / x.qty_count), 5) as qtAVG,
   round(x.qty_stddev, 5)              as qtSTDDEV,
   x.qty_count                         as qtCOUNT,
   round((x.rel_sum / x.rel_count), 5) as relAVG,
   round(x.rel_stddev, 5)              as relSTDDEV,
   x.rel_count                         as relCOUNT,
FROM (SELECT mid,
         SUM(CASE WHEN (mt = "qt") THEN 1 ELSE 0 END)   as qty_count,
         SUM(CASE WHEN (mt = "qt") THEN rt ELSE 0 END)  as qty_sum,
         STD(CASE WHEN (mt = "qt") THEN rt ELSE 0 END)  as qty_stddev
         SUM(CASE WHEN (mt = "rel") THEN 1 ELSE 0 END)  as rel_count,
         SUM(CASE WHEN (mt = "rel") THEN rel ELSE 0 END) as rel_sum,
         STD(CASE WHEN (mt = "rel") THEN rel ELSE 0 END) as rel_stddev
  FROM t_r
  GROUP BY mid) x;

Answer 1

我认为您唯一的问题是ELSE 0。您只需要NULL值，因为它们将被忽略：

SELECT mid                                 as mID,
       round((x.qty_sum / x.qty_count), 5) as qtAVG,
       round(x.qty_stddev, 5)              as qtSTDDEV,
       x.qty_count                         as qtCOUNT,
       round((x.rel_sum / x.rel_count), 5) as relAVG,
       round(x.rel_stddev, 5)              as relSTDDEV,
       x.rel_count                         as relCOUNT,
FROM (SELECT mid,
             SUM( mt = 'qt' )   as qty_count,
             SUM(CASE WHEN mt = 'qt' THEN rt END)  as qty_sum,
             STD(CASE WHEN mt = 'qt' THEN rt END)  as qty_stddev,
             SUM( mt = 'rel' ) as rel_count,
             SUM(CASE WHEN mt = 'rel' THEN rel END) as rel_sum,
             STD(CASE WHEN mt = 'rel' THEN rel END) as rel_stddev
      FROM t_r
      GROUP BY mid
     ) x;

请注意其他一些更改：

我简化了计数逻辑，以删除CASE表达式。它使用MySQL扩展，将布尔值视为数字，其中1为真，0为false。
我用双引号替换了双引号。单引号是字符串的标准分隔符。
我删除了ELSE子句。聚合函数忽略NULL值，因此这可以解决您的问题。

Answer 2

似乎您正在尝试通过该子查询使mysql超越智能。就您而言，您不需要这种额外的复杂性，只需使用带有WHERE子句的简单聚合查询即可，该子句可过滤具有mt =“ qt”的记录。

SELECT 
    mid as mID,
    ROUND(AVG(rt), 5) as qtAVG,
    ROUND(STD(rt), 5) as qtSTDDEV,
    COUNT(*) as qtCOUNT,
FROM t_r
WHERE mt =  "qt"
GROUP BY mt

计算数据透视表内的标准偏差

2 个答案: