如何比较两个表，以计算两个时间段之间的增长率？

Posted 2023-03-31

【中文标题】如何比较两个表，以计算两个时间段之间的增长率？

【英文标题】：How to compare two tables, in order to calculate the growth rate between two time periods?

【发布时间】：2020-09-12 04:14:53

【问题描述】：

我目前拥有的表格如下所示： （数据来自两个不同的表，19921231、19930331）

我要创建的表是这样的（添加了第 5 列）

目标：确定每家银行的存款增长率。 IE。比较上一季度在银行持有的存款（例如 19921231）与最近一个季度的存款（例如 19930331）。然后以百分比的形式查看增加/减少。

这是我目前写的代码：

select 
AL.repdte as `Date`, AL.cert, AL.name, AL.dep as `Deposits`
FROM usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as AL

UNION ALL

select 
AL.repdte as `Date`, AL.cert, AL.name, AL.dep as `Deposits`
FROM usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities as AL

对这个问题的回答建议了这个代码，它有效 但是，由于某种原因，我得到了“NULL”的输出

select al19930331.repdte as `Date`, al19930331.cert, al19930331.name,
       al19930331.dep as Deposits_1993,
       al19921231.dep as Deposits_1992,
       (al19930331.dep - al19921231.dep) / al19921231.dep as grow_rate
from usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities as al19930331 left join
     usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as al19921231
     on al19930331.cert = al19921231.cert and
        al19930331.name = al19921231.name and
        al19921231.repdte = date_add(al19930331.repdte, interval 1 year);

为了隔离“NULL”问题，我将查询简化为最简单的术语，并且能够消除“NULL”问题。

现在我们有两个季度的存款列，返回似乎是正确的输出。

接下来我删除了最后一行代码：

select al19930331.repdte as `Date`, al19930331.cert, al19930331.name,
       al19930331.dep as Deposits_1993,
       al19921231.dep as Deposits_1992,
       (al19930331.dep - al19921231.dep) / al19921231.dep as grow_rate
from usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities as al19930331 left join
     usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as al19921231
     on al19930331.cert = al19921231.cert and
        al19930331.name = al19921231.name and
        al19921231.repdte = date_add(al19930331.repdte, interval 1 year);

删除最后一行代码确实有效。运行代码会产生“除以零”错误。如何消除除以零误差？

select al19930331.repdte as `Date`, al19930331.cert, al19930331.name,
       al19930331.dep as Deposits_1993,
       al19921231.dep as Deposits_1992,
       (al19930331.dep - al19921231.dep) / al19921231.dep as grow_rate
from usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities as al19930331 left join
     usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as al19921231
     on al19930331.cert = al19921231.cert and
        al19930331.name = al19921231.name

【问题讨论】：

你的数据库是什么？如果是 mysql，您使用的是版本 8 还是更高版本？

使用 BigQuery 标准 SQL

您的回复中有一个真实的LAG，但请参阅@GMB 下面的有用答案，看起来是正确的。

【参考方案1】：

您不应将此信息存储在不同的表中。这应该都在同一个表中，使用不同的分区。但是由于名称中间嵌入了日期，我认为您需要使用明确的join：

    select al19930331.repdte as `Date`, al19930331.cert, al19930331.name,
       al19930331.dep as Deposits_1993_0331,
       al19921231.dep as Deposits_1992_1231,
       (al19930331.dep - al19921231.dep) / al19921231.dep as grow_rate
from usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities as al19930331 left join
     usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as al19921231
     on al19930331.cert = al19921231.cert and 
     al19921231.repdte = date_add(al19930331.repdte, interval 1 quarter);

将数据放在一个表中会更简单。

【讨论】：

在 Deposits_1992 和 Growth_rate 上接收 NULL。选择al19930331.repdte如Date，al19930331.cert，al19930331.name，al19930331.dep如Deposits_1993_0331，al19921231.dep如Deposits_1992_1231，（al19930331.dep - al19921231.dep）/ al19921231.dep从usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities grow_rate作为al19930331 left join usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities as al19921231 on al19930331.cert = al19921231.cert 和 al19921231.repdte = date_add(al19930331.repdte, 间隔 1 季度);

@user2755660 。 . .那么您在上一年没有匹配的值。也许名称不匹配。

将查询配对以消除 Null。现在正在返回两个季度的存款值。现在只需要推理。想知道这是否可能是除法/零问题。选择al19930331.repdte为Date，al19930331.cert，al19930331.name，al19930331.dep为Deposits_1993_0331，al19921231.dep为Deposits_1992_1231，从usa_fdic_call_reports_1993.All_Reports_19930331_Assets_and_Liabilities作为al19930331左连接usa_fdic_call_reports_1992.All_Reports_19921231_Assets_and_Liabilities作为al19921231上al19930331.cert = al19921231.cert

【参考方案2】：

你可以使用lag():

select
    t.*,
    100.0 
        * (deposits - lag(deposits) over(partition by cert order by depdte)) 
        / lag(deposits) over(partition by cert order by depdte) as growth_percent
from mytable t