用MySQL计算中位数最简单(希望不会太慢)的方法是什么?我已经使用AVG(x)来寻找平均值，但我很难找到一个简单的方法来计算中位数。现在，我将所有的行返回到PHP，进行排序，然后选择中间的行，但是肯定有一些简单的方法可以在一个MySQL查询中完成它。
示例数据:
id | val
--------
1 4
2 7
3 2
4 2
5 9
6 8
7 3
对val排序得到2 2 3 4 7 8 9，因此中位数应该是4，而SELECT AVG(val) == 5。
当前回答
下面的查询对于奇数行和偶数行都非常有效。在子查询中，我们正在寻找前后行数相同的值。对于奇数行的情况，having子句的值将为0(前后相同的行数将抵消符号)。
类似地，对于偶数行，having子句对于两行(中间的两行)的计算结果为1，因为它们(总的来说)前后的行数相同。
在外层查询中，我们将平均出单个值(奇数行)或(偶数行2个值)。
select avg(val) as median
from
(
select d1.val
from data d1 cross join data d2
group by d1.val
having abs(sum(sign(d1.val-d2.val))) in (0,1)
) sub
注意:如果你的表有重复的值，上面的having子句应该更改为下面的条件。在这种情况下，可能有一些值超出了原来的可能性(0,1)下面的条件将使这个条件动态，并在重复的情况下工作。
having sum(case when d1.val=d2.val then 1 else 0 end)>=
abs(sum(sign(d1.val-d2.val)))
其他回答
在某些情况下，中位数的计算如下:
“中位数”是数字列表中按值排序时的“中间”值。对于偶数集，中位数是两个中间值的平均值。 我为此创建了一个简单的代码:
$midValue = 0;
$rowCount = "SELECT count(*) as count {$from} {$where}";
$even = FALSE;
$offset = 1;
$medianRow = floor($rowCount / 2);
if ($rowCount % 2 == 0 && !empty($medianRow)) {
$even = TRUE;
$offset++;
$medianRow--;
}
$medianValue = "SELECT column as median
{$fromClause} {$whereClause}
ORDER BY median
LIMIT {$medianRow},{$offset}";
$medianValDAO = db_query($medianValue);
while ($medianValDAO->fetch()) {
if ($even) {
$midValue = $midValue + $medianValDAO->median;
}
else {
$median = $medianValDAO->median;
}
}
if ($even) {
$median = $midValue / 2;
}
return $median;
返回的$中位数将是所需的结果:-)
通常，我们不仅需要为整个表计算Median，还需要为与ID相关的聚合计算Median。换句话说，计算表中每个ID的中位数，其中每个ID有许多记录。(良好的性能和工作在许多SQL +修复偶数和赔率的问题，更多关于不同的中值方法的性能https://sqlperformance.com/2012/08/t-sql-queries/median)
SELECT our_id, AVG(1.0 * our_val) as Median
FROM
( SELECT our_id, our_val,
COUNT(*) OVER (PARTITION BY our_id) AS cnt,
ROW_NUMBER() OVER (PARTITION BY our_id ORDER BY our_val) AS rn
FROM our_table
) AS x
WHERE rn IN ((cnt + 1)/2, (cnt + 2)/2) GROUP BY our_id;
希望能有所帮助
如果这是MySQL，现在有窗口函数，你可以这样做(假设你想四舍五入到最接近的整数-否则只需将round替换为CEIL或FLOOR或其他什么)。下面的解决方案适用于表，无论表的行数是偶数还是奇数:
WITH CTE AS (
SELECT val,
ROW_NUMBER() OVER (ORDER BY val ASC) AS rn,
COUNT(*) OVER () AS total_count
FROM data
)
SELECT ROUND(AVG(val)) AS median
FROM CTE
WHERE
rn BETWEEN
total_count / 2.0 AND
total_count / 2.0 + 1;
I think some of the more recent answers on this thread were already getting at this approach, but it also seemed like people were overthinking it, so consider this an improved version. Regardless of SQL flavor, there is no reason anyone should be writing a huge paragraph of code with multiple subqueries just to get the median in 2021. However, please note that the above query only works if you're asked to find the median for a continuous series. Of course, regardless of row number, sometimes people do make a distinction between what is referred to as the Discrete Median and what is referred to as the Interpolated Median for a continuous series.
如果你被要求为一个离散级数找到中位数，而表的行数是偶数，那么上面的解决方案就不适合你，你应该恢复使用其他解决方案之一，比如TheJacobTaylor的。
下面的第二个解决方案是对TheJacobTaylor的稍微修改的版本，其中我显式地声明了CROSS JOIN。这个方法也适用于行数为奇数的表，不管你是被要求求连续序列的中位数还是离散序列的中位数，但我特别会在被要求求离散序列的中位数时使用这个方法。否则，使用第一种解决方案。这样，您就永远不必考虑数据是包含“偶数”还是“奇数”个数的数据点。
SELECT x.val AS median
FROM data x
CROSS JOIN data y
GROUP BY x.val
HAVING SUM(SIGN(1 - SIGN(y.val - x.val))) = (COUNT(*) + 1) / 2;
最后，你可以在PostgreSQL中使用内置函数轻松做到这一点。这里有一个很好的解释，以及关于离散中位数和插值中位数的有效总结。
https://leafo.net/guides/postgresql-calculating-percentile.html#calculating-the-median
我建议一个更快的方法。
获取行数:
SELECT CEIL(COUNT(*)/2);
然后取排序子查询的中间值:
SELECT max(val) FROM (SELECT val FROM data ORDER BY val limit @middlevalue);
我用5x10e6的随机数数据集进行了测试，它将在10秒内找到中位数。
按维度分组的中位数:
SELECT your_dimension, avg(t1.val) as median_val FROM (
SELECT @rownum:=@rownum+1 AS `row_number`,
IF(@dim <> d.your_dimension, @rownum := 0, NULL),
@dim := d.your_dimension AS your_dimension,
d.val
FROM data d, (SELECT @rownum:=0) r, (SELECT @dim := 'something_unreal') d
WHERE 1
-- put some where clause here
ORDER BY d.your_dimension, d.val
) as t1
INNER JOIN
(
SELECT d.your_dimension,
count(*) as total_rows
FROM data d
WHERE 1
-- put same where clause here
GROUP BY d.your_dimension
) as t2 USING(your_dimension)
WHERE 1
AND t1.row_number in ( floor((total_rows+1)/2), floor((total_rows+2)/2) )
GROUP BY your_dimension;
