Question

我在Google的BigQuery中有以下SQL查询：

SELECT
  sensor,
  date,
  time,
  perc,
  PERCENTILE_CONT(perc,
    0.2) OVER(PARTITION BY sensor, date) AS percentile20_day,
  PERCENTILE_CONT(perc,
    0.8) OVER(PARTITION BY sensor, date) AS percentile80_day
FROM
  sensordata
ORDER BY
  Sensor,
  date,
  time

现在，我要添加的另一项计算是两个百分点之间的平均值。在此特定情况下，整天所有20％至80％百分数之间的平均值。不过，我不想仅限于一天，而可能将其更改为时间窗口（例如，通过添加另一个分区）。

到目前为止，我还无法完成这项工作。我不明白如何添加它。

Answer 1

您需要使用子查询：

SELECT sensor, date, time, perc, percentile20_day, percentile80_day,
       AVG(CASE WHEN perc >= percentile20_day AND perc <= percentile90_day THEN perc END) OVER () as avg_20_80,
FROM (SELECT sd.*
            PERCENTILE_CONT(perc, 0.2) OVER (PARTITION BY sensor, date) AS percentile20_day,
            PERCENTILE_CONT(perc, 0.8) OVER (PARTITION BY sensor, date) AS percentile80_day
      FROM sensordata sd
     ) sd
ORDER BY Sensor, date, time;

Answer 2

以下是用于BigQuery标准SQL

 Fetched in submodule path 'B', but it did not contain 3ffeb07de8634a3540d185cbd16f76ebee704f7d. Direct fetching of that commit failed.

Answer 3

尝试一下：

SELECT date, AVG(CASE
    WHEN sensor < PERCENTILE_CONT(perc, 0.2) OVER(PARTITION BY sensor, date) THEN NULL
    WHEN sensor > PERCENTILE_CONT(perc, 0.8) OVER(PARTITION BY sensor, date) THEN NULL
    ELSE sensor
    ) AS AvgSensorBTW20thAnd80thPercentile
FROM sensordata
GROUP BY date

注意：未经测试！

根据MSDN文档AVG，函数将忽略NULL。

如何计算20％和80％之间的平均值？

3 个答案: