How to calculate rolling 3 months average to replace missing data?

[Ankita_]

I have the following data, where some of the quantities are missing. I want to replace them by a rolling 3 month average of previous 3 months. I've tried to explain what I'm trying to achieve here- 

ID	Month	Qty	Rolling average	Rolling average value
1	Jan	2	-	-
1	Feb	3	2	2
1	Mar	2	average(3,2)	2.5
1	Apr	6	average(2,3,2)	2.333333333
1	May	4	average(6,2,3)	3.666666667
1	Jun	5	AVERAGE(4,6,2)	4
1	Jul	.	average(5,4,6)	5
1	Aug	.	average(5,5,4)	4.666666667
1	Sep	.	average(4.666667,5,5)	4.888888889
1	Oct	.	average(4.8888889,4.666667,5)	4.851851852
1	Nov	.	average(4.8518519, 4.888889, 4.666667)	4.802469136
1	Dec	.	average(4.8024691, 4.8518519, 4.8888889)	4.847736626
2	Jan	5
2	Feb	1
2	Mar	4
2	Apr	3
2	May	4
2	Jun	6
2	Jul	.
2	Aug	.
2	Sep	.
2	Oct	.
2	Nov	.
2	Dec	.

Can anyone please help on how to do this in SAS? 

[PeterClemmensen · Accepted solution]

Try this and replace as you like.

data have;
input ID Month $ Qty;
datalines; 
1 Jan 2 
1 Feb 3 
1 Mar 2 
1 Apr 6 
1 May 4 
1 Jun 5 
1 Jul . 
1 Aug . 
1 Sep . 
1 Oct . 
1 Nov . 
1 Dec . 
2 Jan 5 
2 Feb 1 
2 Mar 4 
2 Apr 3 
2 May 4 
2 Jun 6 
2 Jul . 
2 Aug . 
2 Sep . 
2 Oct . 
2 Nov . 
2 Dec . 
;

data want;
    array lag[0:2] _temporary_;        
    call missing(of lag[*]);        

    do _N_ = 1 by 1 until (last.id);      
        set have;
        by id;

        output;

        lag[mod(_N_, 3)] = coalesce(Qty, avg);            
        avg = mean(of lag[*]);
    end;
run;

Result:

ID  Month  Qty  avg
1   Jan    2    .
1   Feb    3    2
1   Mar    2    2.5
1   Apr    6    2.3333333333
1   May    4    3.6666666667
1   Jun    5    4
1   Jul    .    5
1   Aug    .    4.6666666667
1   Sep    .    4.8888888889
1   Oct    .    4.8518518519
1   Nov    .    4.8024691358
1   Dec    .    4.8477366255
2   Jan    5    .
2   Feb    1    5
2   Mar    4    3
2   Apr    3    3.3333333333
2   May    4    2.6666666667
2   Jun    6    3.6666666667
2   Jul    .    4.3333333333
2   Aug    .    4.7777777778
2   Sep    .    5.037037037
2   Oct    .    4.7160493827
2   Nov    .    4.8436213992
2   Dec    .    4.865569273

[Patrick]

Depending on your real data this discussion might provide all the answers you need.