July 1, 2009 at 9:31 am
I want to return all values that represent x% of the total values within a table. Here's some test data:
create table #temp (customer varchar(15),number int)
insert into #temp select'A',10
insert into #temp select'B',20
insert into #temp select'C',17
insert into #temp select'D',18
insert into #temp select'E',30
insert into #temp select'F',40
insert into #temp select'G',10
insert into #temp select'H',20
insert into #temp select'I',17
insert into #temp select'J',18
insert into #temp select'K',30
insert into #temp select'L',40
select top 60 percent * from #temp
order by number desc
drop table #temp
When this is executed 60% of the total lines are returned but what I want is for it to return those rows whose 'number' values represent 60% of the total when added. So using my test data, 60% of the total would be represented by customers F, L, K and E because the sum of all the rows is 233 and the sum of F, L, K and E is 140 = 60%
Thanks
David
July 1, 2009 at 10:34 am
July 1, 2009 at 10:47 am
This is a weird one...
would something like this do the trick?
declare @Total decimal(18,2)
select @Total = sum(number)
from #temp
create table #temp2
(
customer varchar(15),
number int,
Ranking int
)
insert #temp2
select customer,
Number,
ranking = row_number() over(order by number desc)
from #temp
select customer,
number,
Running,
pct = round((cast(Running as decimal(10,2))/ cast(@Total as decimal(10,2)))*100, 0)
from
(
select customer,
number,
Ranking = ntile(12) over(order by number desc),
Running = (select sum(number) from #temp2 t1
where t1.Ranking <= t.Ranking)
from #temp2 t
)sub
where round((cast(Running as decimal(10,2))/ cast(@Total as decimal(10,2)))*100, 0) <= 60
drop table #temp2
This mess first orders the customers by Number Desc. Then calculates the running total and percentage of the running total against the total total, and returns those rows that add up to close to 60% without going over ("you're the next contestant on the Price is Right"...)
This is based on an inference that you want to return the top 60% of customers by number.
This works fine on a small set, but doubt it would be pretty on a large set of data...lots of table scans, sorts and temp table weirdness...
/*****************
If most people are not willing to see the difficulty, this is mainly because, consciously or unconsciously, they assume that it will be they who will settle these questions for the others, and because they are convinced of their own capacity to do this. -Friedrich August von Hayek
*****************/
July 1, 2009 at 10:52 am
I wasn't sure if you were looking for all combinations of Customers that meet that percentage or just starting with the min number or.???
At anyrate, I just used a simple running total solution to get some results. Hopefully, it'll help:DECLARE @Temp table (customer varchar(15),number int)
insert into @Temp select'A',10
insert into @Temp select'B',20
insert into @Temp select'C',17
insert into @Temp select'D',18
insert into @Temp select'E',30
insert into @Temp select'F',40
insert into @Temp select'G',10
insert into @Temp select'H',20
insert into @Temp select'I',17
insert into @Temp select'J',18
insert into @Temp select'K',30
insert into @Temp select'L',40
DECLARE @Sum INT
SELECT @Sum = SUM(number)
FROM @Temp
DECLARE @Percent FLOAT
SET @Percent = @Sum * .6
SELECT
customer, Number, RunningTotal
FROM
(
SELECT
customer,
number,
(
SELECT SUM(Number)
FROM
(
SELECT *,ROW_NUMBER() OVER (ORDER BY Number, Customer) AS RowNum
FROM @Temp
) AS A
WHERE RowNum <= T.RowNum
) AS RunningTotal
FROM
(SELECT *,ROW_NUMBER() OVER (ORDER BY Number, Customer) AS RowNum FROM @Temp) T
) AS D
WHERE
RunningTotal <= @Percent -- Could add on some sort of varience if needed
ORDER BY
RunningTotal
July 1, 2009 at 1:26 pm
Thanks all, I'm out of the office so I'll test your solutions tomorrow.
Just to clarify, what I wanted was to first order the rows in the temp table by number (desc) and then count 60% from the top downwards. Once 60% has been reached, or as close to, then stop. I see that you've already guessed that though!
July 2, 2009 at 9:35 am
Ok, I have had a chance to test out your solutions and am glad to say that after a bit of fiddling it is returning the data I need! We have a mixed environment here where some databases are not SQL 2005 compatible so the row numbering had to be done "the old way". Nevertheless, it is working now so thank you for your help.
December 5, 2012 at 8:06 am
Hi David:
Could you post your final solution? I have found out that two suggestions gave two different results and I would like to know the right one you have finally used.
Thanks,
Adrian
December 13, 2012 at 4:28 am
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply