在 MySQL 中,如何计算一组数据的中位数?

要得到一组数据的中位数(例如某个地区或某家公司的收入中位数),我们一般要将这一任务细分为 3 个小任务:

  1. 将数据排序,并给每一行数据给出其在所有数据中的排名;
  2. 找出中位数的排名数字;
  3. 找出中间排名对应的值;

下面以某公司员工月收入为例,示例 MySQL 的一些复杂语句的使用。

方法一

创建测试表

首先创建一个收入表,建表语句为:

CREATE TABLE IF NOT EXISTS `employee` (
  `id`     INT                  AUTO_INCREMENT PRIMARY KEY,
  `name`   VARCHAR(10) NOT NULL DEFAULT '',
  `income` INT         NOT NULL DEFAULT '0'
)
  ENGINE = InnoDB
  DEFAULT CHARSET = utf8;

INSERT INTO `employee` (`name`, `income`)
VALUES ('麻子', 20000);
INSERT INTO `employee` (`name`, `income`)
VALUES ('李四', 12000);
INSERT INTO `employee` (`name`, `income`)
VALUES ('张三', 10000);
INSERT INTO `employee` (`name`, `income`)
VALUES ('王二', 16000);
INSERT INTO `employee` (`name`, `income`)
VALUES ('土豪', 40000);

完成任务1

将数据排序,并给每一行数据给出其在所有数据中的排名:

SELECT t1.name, t1.income, COUNT(*) AS rank
FROM employee AS t1,
     employee AS t2
WHERE t1.income < t2.income
   OR (t1.income = t2.income AND t1.name <= t2.name)
GROUP BY t1.name, t1.income
ORDER BY rank;

查询结果为:

name income rank
土豪 40000 1
麻子 20000 2
王二 16000 3
李四 12000 4
张三 10000 5

完成小任务2

找出中位数的排名数字:

SELECT (COUNT(*) + 1) DIV 2 as rank
FROM employee;

查询结果为:

rank
3

完成小任务3

SELECT income AS median
FROM (SELECT t1.name, t1.income, COUNT(*) AS rank
      FROM employee AS t1,
           employee AS t2
      WHERE t1.income < t2.income
         OR (t1.income = t2.income AND t1.name <= t2.name)
      GROUP BY t1.name, t1.income
      ORDER BY rank) t3
WHERE rank = (SELECT (COUNT(*) + 1) DIV 2 FROM employee)

查询结果为:

median
16000

至此,我们就找到了如何从一组数据中获得中位数的方法。

方法二

下面,来介绍另外一种优化排名语句的方法。

我们都知道如何给一组数据做排序操作,在本例中,实现方法如下:

SELECT name, income
FROM employee
ORDER BY income DESC

查询结果为:

name income
土豪 40000
麻子 20000
王二 16000
李四 12000
张三 10000

那我们可不可以更进一步,对查询出的结果加一列,这一列的数据为排名呢?

我们可以通过3个自定义变量的方法来实现这一目标:

第一个变量用来记录当前行数据的收入
第二个变量用来记录上一行数据的收入
第三个变量用来记录当前行数据的排名

SET @curr_income := 0;
SET @prev_income := 0;
SET @rank := 0;

SELECT `name`,
       @curr_income := income                                      AS income,
       @rank := if(@prev_income != @curr_income, @rank + 1, @rank) AS rank,
       @prev_income := @curr_income                                AS dummy
FROM employee
ORDER BY income DESC

查询结果如下:

name income rank dummy
土豪 40000 1 40000
麻子 20000 2 20000
王二 16000 3 16000
李四 12000 4 12000
张三 10000 5 10000

然后再找出中位数的排名数字,进一步找出收入的中位数:

SET @curr_income := 0;
SET @prev_income := 0;
SET @rank := 0;

SELECT income AS median
FROM (SELECT `name`,
             @curr_income := income                                      AS income,
             @rank := if(@prev_income != @curr_income, @rank + 1, @rank) AS rank,
             @prev_income := @curr_income                                AS dummy
      FROM employee
      ORDER BY income DESC) AS t1
WHERE t1.rank = (SELECT (COUNT(*) + 1) DIV 2 FROM employee)

查询结果为:

median
16000

至此,我们找了两种方法来解决中位数的问题。撒花。

本作品采用《CC 协议》,转载必须注明作者和本文链接
本帖由系统于 4年前 自动加精
讨论数量: 3

这里name 为中文怎么比较的?

SELECT t1.name, t1.income, COUNT(*) AS rank
FROM employee AS t1,
     employee AS t2
WHERE t1.income < t2.income
   OR (t1.income = t2.income AND t1.name <= t2.name)
GROUP BY t1.name, t1.income
ORDER BY rank;
5年前 评论

@lovecn 根据编码格式,一般是 unicode 编码的顺序来确定大小。

5年前 评论

任务 1 可以直接用下面的代码吗?

select t1.name, t1.income, count(*) as rank
from employee as t1,
employee as t2 
where t1.`income` <= t2.income
group by t1.name,t1.income
order by rank;
4年前 评论

讨论应以学习和精进为目的。请勿发布不友善或者负能量的内容,与人为善,比聪明更重要!