如何过滤掉特定列下具有重复值的数据？

Question

我有一个包含调查结果的表格：

submitter   issue       q1  q2  q3  q4  q5

mike         11557      4   3   4   5   1
mark         13554      5   5   5   5   5
luke         15110      1   1   1   1   1
luke         15110      1   1   1   1   1
donald       16900      4   2   2   4   5
joe          11562      5   5   5   5   5
joe          11562      5   5   5   5   5
sam          12485      2   3   4   3   4
sam          12485      2   3   4   3   4
sam          12485      2   3   4   3   4

我希望能够过滤掉多个提交并仅计算其中的 1 个。 有些人提交了 3 或 4 次。

我知道如何知道调查提交了多少次以及由谁提交：​​

SELECT
    submitter
    ,issue
    ,COUNT(*) as '# of times Survey submitted'

FROM
    Survey

GROUP BY
    submitter, issue

HAVING
    COUNT(*) > 1

但是，我不确定如何使用此查询来过滤掉多个提交。

我正在处理的当前查询是：

SELECT 'Question #1' as 'Survey Question'
,CAST(CAST(SUM(q1) AS float)/COUNT(q1) AS decimal (4,2)) as 'Average Score'

FROM Survey
WHERE COALESCE(q1,q2,q3,q4,q5) IS NOT NULL

UNION ALL

SELECT 'Question #2' as 'Survey Question'
,CAST(CAST(SUM(q2) AS float)/COUNT(q2) AS decimal (4,2)) as 'Average Score'
FROM Survey
WHERE COALESCE(q1,q2,q3,q4,q5) IS NOT NULL

UNION ALL

etc...

期望的结果是：（注意：这个结果集不准确。只是我想要的格式。）

Survey Question Average Score
Question #1      4.58
Question #2      4.80
Question #3      4.60
Question #4      4.59
Question #5      4.64

谁能提供线索？

非常感谢！

Answer 1

我认为我的数学计算是正确的，但我的结果与您的结果不完全相符。你确定你想要的结果是正确的吗？

DECLARE @yourTable TABLE (submitter VARCHAR(10), Issue INT, q1 TINYINT, q2 TINYINT,q3 TINYINT, q4 TINYINT,q5 TINYINT);
INSERT INTO @yourTable
VALUES  ('mike',11557,4,3,4,5,1),
        ('mark',13554,5,5,5,5,5),
        ('luke',15110,1,1,1,1,1),
        ('luke',15110,1,1,1,1,1),
        ('donald',16900,4,2,2,4,5),
        ('joe',11562,5,5,5,5,5),
        ('joe',11562,5,5,5,5,5),
        ('sam',12485,2,3,4,3,4),
        ('sam',12485,2,3,4,3,4),
        ('sam',12485,2,3,4,3,4);

WITH CTE_Distinct
AS
(
    SELECT DISTINCT *
    FROM @yourTable  --just change this to your actual table name.
)

SELECT  REPLACE(question,'q','Question #')   AS [Survey Question],
        CAST(AVG(val * 1.0) AS DECIMAL(4,2)) AS [Average Score]
FROM CTE_Distinct
UNPIVOT
(
    val FOR question IN (q1,q2,q3,q4,q5)
) unpvt
GROUP BY question

结果：

Survey Question     Average Score
-------------------- ---------------------------------------
Question #1          3.50
Question #2          3.17
Question #3          3.50
Question #4          3.83
Question #5          3.50

Answer 2

WITH TestData AS (
    SELECT *
    FROM (VALUES
        ('Mike', 11557, 4, 3, 4, 5, 1)
      , ('Mark', 13554, 5, 3, 5, 5, 5)
      , ('Luke', 15110, 1, 1, 1, 1, 1)
      , ('Luke', 15110, 1, 1, 1, 1, 1)
      , ('Donald', 16900, 4, 2, 2, 4, 5)
      , ('Joe', 11562, 5, 5, 5, 5, 5)
      , ('Joe', 11562, 5, 5, 5, 5, 5)
      , ('Sam', 12485, 2, 3, 4, 3, 4)    
      , ('Sam', 12485, 2, 3, 4, 3, 4)    
      , ('Sam', 12485, 2, 3, 4, 3, 4)    
    ) A (Submitter, Issue, Q1, Q2, Q3, Q4, Q5)
)
SELECT SurveyQuestion
     , AverageScore = AVG(QuestionAnswer * 1.) -- Change the math here if this isn't what you want 
FROM (    
    SELECT A.Submitter
         , A.Issue
         , B.SurveyQuestion
         , B.QuestionAnswer
         , RowNum = ROW_NUMBER() OVER(PARTITION BY A.Submitter, A.Issue, B.SurveyQuestion ORDER BY (SELECT NULL)) -- Replace ORDER BY (SELECT NULL) with something more meaningful if you can
    FROM TestData A
    CROSS APPLY(VALUES -- Unpivot
        ('Question #1', A.Q1)
      , ('Question #2', A.Q2)
      , ('Question #3', A.Q3)
      , ('Question #4', A.Q4)
      , ('Question #5', A.Q5)
    ) B (SurveyQuestion, QuestionAnswer)
    WHERE B.SurveyQuestion IS NOT NULL
) A
WHERE RowNum = 1
GROUP BY SurveyQuestion;

Answer 3

我认为您可以应用的第一个解决方案是：选择提交者和问题以及每个提交者给出的每个答案的最大值：

    select submitter, issue, 
       (select max(q1) 
           from survey 
          where submitter = parent.submitter 
            and issue = parent.issue) as q1,
       (select max(q2) 
           from survey 
          where submitter = parent.submitter 
            and issue = parent.issue) as q2,
       (select max(q3) 
           from survey 
          where submitter = parent.submitter 
            and issue = parent.issue) as q3,
       (select max(q4) 
           from survey 
          where submitter = parent.submitter 
            and issue = parent.issue) as q4,
       (select max(q5) 
           from survey 
          where submitter = parent.submitter 
            and issue = parent.issue) as q5
from survery as parent
group by submitter, issue; 

但是这个解决方案的问题是它给出了例如每个问题的最佳答案，这可能不是所需的输出。

另一种方法是为每个寄存器添加一个 id：

alter table survery add id bigint auto_increment;

使用将每一行标记为不同的 id，这是另一条 keetle 的鱼。选择要简单得多：

    select *
      from survey
      where (submitter, issue, id ) in 
      (
    select submitter, issue, max(id)
      from survey 
     group by submitter, issue);

内部选择（具有分组依据的选择）标识您想要获取的 id，第二个选择检索所有信息：提交者、id 和答案。您可以将它与 max() 一起使用来检索最后一个答案作为 good 答案，或者您可以将它与 min() 一起使用来检索第一个答案。

更新

抱歉，我没有看到您提出的“平均”要求。如果您想要一个平均值而不是答案，我谦虚地推荐第二种方法。选择将是：

    select avg(q1) as avg_q1, 
           avg(q2) as avg_q2, 
           ....
      from survey
      where (submitter, issue, id ) in 
      (
    select submitter, issue, max(id)
      from survey 
     group by submitter, issue);