Sum column based on row count

[Eric21]

Hello,

I have searched all over and can't seem to find an answer for this. For each row, I am attempting to sum the values of the next x rows (3 in the example below) for each item.

Example: Item 1, row count for aggregation = 3, next 3 rows countain 10 in each row which equals a sum of 30.  It only sums for the item that corresponds to the row that the sum calculation is on (similar to a group by). 

The example below is far simpler than my real data which contains different values for sales for each row.

I am assuming that this would be an iterating do until loop of some sort but I can't seem to think of a strategy.  Thanks for all of your insight!

Item	Sales	SumRolling
1	10	30
1	10	30
1	10	30
1	10	30
1	10	30
1	10	30
1	10	20
1	10	10
2	20	60
2	20	60
2	20	60
2	20	60
2	20	60
2	20	40
2	20	20

[Reeza]

Do you have SAS ETS? If so, a CONVERT statement is your best bet.

If not, the quickest way, IMO is to reverse the data, use the LAG function and then reverse the data order again. Assuming you can do the re-ordering yourself, here's how the data step may look. 

data want; 
by item sales;

l1=lag(sales);
l2=lag2(sales);
l3=lag3(sales);

if first.item then count=1; else count+1;

if count > 3 then do;
sum = sum(l1, l2, l3);
end;

run;

[Ksharp]

data have;
infile cards expandtabs;
input Item	Sales;
cards;
1	10	30
1	10	30
1	10	30
1	10	30
1	10	30
1	10	30
1	10	20
1	10	10
2	20	60
2	20	60
2	20	60
2	20	60
2	20	60
2	20	40
2	20
;
run;
data have;
 set have;
 by item;
 if first.item then n=0;
 n+1;
run;
proc sql;
select a.*,(select sum(sales) from have where item=a.item and 
n between a.n and a.n+2) as rolling_sum
 from have as a;
quit;