Google bigquery BigQuery中的percent_rank（），条件为仅包含某些行

我以前发过。此问题的解决方案适用于分析函数

rank（）

，但不适用于

percent\u rank（）

。为了演示，我有以下虚拟表格：

with 
  table as (
    select 'a' as category, 1 as num, 15 as num2, 7 as cutoff union all 
    select 'a' as category, 2 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 3 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 4 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 5 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 6 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 7 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 8 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 9 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 10 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 11 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 12 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 13 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 14 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 15 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 16 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 17 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 18 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 19 as num, 15 as num2, 7 as cutoff union all
    select 'a' as category, 20 as num, 5 as num2, 7 as cutoff union all
    select 'a' as category, 21 as num, 5 as num2, 7 as cutoff 
  )

num

列需要

percent\u rank（）

。但是，百分位排名只应考虑

num2>截止值

的行。我尝试了以下两种方法来计算百分位数，并给出了结果：

select
  *,
  if(num2 >= cutoff,
      percent_rank() over(
        partition by category
        order by num
      ), null) as pctile1,
  if(num2 >= cutoff,
      percent_rank() over(
        partition by category
        order by if (num2 >= cutoff, num, null) ASC
      ), null) as pctile2
from table
order by num asc

pctile1

和

pctile2

都不正确。要说明为什么会出现这种情况，请查看第10行，该行具有

pctile1==0.45

和

pctile2==0.60

。然而，在合格值中，这应该是一个较低的百分位数。只有2个符合条件的值低于

num==10

（即1和2），而10以上的许多值符合条件（11-19）。给定

num==10

和

cutoff

值，正确的

num==10的百分位数应该接近30%，因为
10
是11个限定值中的第三个最低值

请注意，我不应筛选表以删除我未

percent_rank（）“覆盖”的行，因为我需要保留这些行

编辑

---

我不知道如何缩小图像大小，但我目前正在尝试这样做。

我只想使用下面的选项

#standardSQL
SELECT *,
  PERCENT_RANK() OVER(PARTITION BY category ORDER BY num) AS pctile
FROM table WHERE num2 >= cutoff
UNION ALL
SELECT *, NULL
FROM table WHERE num2 < cutoff
-- ORDER BY num

在我看来，上面的内容很容易阅读，但下面的内容很可能是你想要的

SELECT *,
  IF(num2 >= cutoff, 
    PERCENT_RANK() OVER(PARTITION BY IF(num2 >= cutoff, category, NULL) ORDER BY num), 
    NULL) AS pctile
FROM table
-- ORDER BY num

---

显然，与上面的结果相同

ahh，我在

orderby

中使用了

if

条件，而不是

分区

。谢谢分享/更正。

SELECT *,
  IF(num2 >= cutoff, 
    PERCENT_RANK() OVER(PARTITION BY IF(num2 >= cutoff, category, NULL) ORDER BY num), 
    NULL) AS pctile
FROM table
-- ORDER BY num