计算 SQL Server 中的平均值

Question

I want to do something very simple but I'm obviously missing a trick! I want to get an average of average values but I want to include the weighting of the original average calculation. I'll use a stripped back version of what I'm attempting to do.

So let's say I have the following table

Product     date        RunInterval    AvgDuration_secs   Executions
--------------------------------------------------------------------
   A        29/12/19            1               1           100
   A        29/12/19            2               2            10

What I want to find out is what the average duration was for Product A on 29/12. All the things I've tried so far are giving me an average of 1.5 secs ie it's adding together the duration of 1 & 2 secs (3) and dividing by number of rows (2) to give 1.5. What I want to get to is to have the average but taking into account how often it runs so (100*1) + (10*2) / 110 = 1.09 secs. I've tried various attempts with GROUP BY statements and CURSORS but not getting there.

I'm evidently tackling it the wrong way! Any help welcome :-)

Answer 1

You can do it like this:

select product, date,
  round(1.0 * sum([Executions] * [AvgDuration_secs]) / sum([Executions]), 2) result
from tablename
group by product, date

I'm not sure if you want RunInterval or AvgDuration_secs in the 1st sum.
See the demo .
Results:

> product | date      | result        
> :------ | :---------| :-----
> A       | 29/12/2019| 1.09

Answer 2

If you got those results from a query or view that select from some table, then grouped by Product & date & RunInterval.

Then you could simply run a query that on that table that groups only by the Product & date.

An example:

 -- -- Sample data -- CREATE TABLE sometable ( Product varchar(30), ExecutionDatetime datetime, RunInterval int ); WITH RCTE_NUMS AS ( SELECT 1 AS n UNION ALL SELECT n+1 FROM RCTE_NUMS WHERE n < 110 ) INSERT INTO sometable (Product, ExecutionDatetime, RunInterval) SELECT 'A' p, DATEADD(minute,n*12,'2019-12-29 01:00:00') dt, IIF(n<=100,1,2) ri FROM RCTE_NUMS OPTION (MAXRECURSION 1000);

\n110 rows affected \n

select Product, cast(ExecutionDatetime as date) as [Date], AVG(1.0*RunInterval) AS AvgDuration_secs, COUNT(*) AS Executions from sometable t group by Product, cast(ExecutionDatetime as date) ORDER BY Product, [Date]

\nProduct | Date | AvgDuration_secs |  Executions\n:------ |  :------------------ |  :--------------- |  ---------: \nA | 29/12/2019 00:00:00 |  1.090909 |  110 \n

db<>fiddle here