Solved
New Contributor
Posts: 4

# Calculating a rolling sum for the last 30 days

[ Edited ]

Hi all,

I have a programming question regarding the calculation of a rolling sum for a range of variables. To simplify the question I have only included two variables to be calculated. The sum must be calculated for each customer represented in the data.

DATA = work.ydelser

PERSON ID         DATE                 TOTAL      PROD_A          ...

A                          08-01-2014         10                 ...

A                          13-01-2014         12                 ...

A                          23-01-2014         11

A                          12-02-2014         17

B                          07-01-2014         26

B                          17-01-2014         50

B                          25-02-2014         10

B                          03-03-2014         10

The result I would like to have is:

DATA = work.ydel_sum

PERSON ID         DATE                 TOTALsum        PROD_A

A                          08-01-2014         10                         ...

A                          13-01-2014         22

A                          23-01-2014         33

A                          12-02-2014         22

B                          07-01-2014         26

B                          17-01-2014         76

B                          25-02-2014         10

B                          03-03-2014         20

The computed dataset does not have to include the original variable. The date is in "date format".

I will use the code in a code-program in Enterprise Guide, but if anyone knows how to solve the question by Enterprise functions, please let me know.

I have seen a privious topic on this matter, but I cannot figure out how to use the "cards" function.

Accepted Solutions
Solution
‎11-26-2015 01:54 AM
Super User
Posts: 9,599

## Re: Calculating a rolling sum for the last 30 days

Here is code which will do a rolling 30 day sum for you.  The datalines or cards is just used to create the test data (I have left in to show you).  I do not understand why the fourth A row in your example out data is 22 though, that doesn't make sense with the logic.

data have;
informat date ddmmyy10.;
format date date9.;
input person_id \$ date total;
datalines;
A 08-01-2014 10
A 13-01-2014 12
A 23-01-2014 11
A 12-02-2014 17
;
run;

proc sql;
create table WANT as
select  PERSON_ID,
DATE,
TOTAL,
sum(TMP) as ROLLING_SUM
from    (
select  A.*,
B.TOTAL as TMP
from    HAVE A
left join HAVE B
on      A.PERSON_ID=B.PERSON_ID
and     A.DATE-30 <= B.DATE <= A.DATE
)
group by PERSON_ID,
DATE,
TOTAL;
quit;

What this does is join all records where date is within 30 days to the original data, this is the subquery.  That data is then grouped by the original variables and a sum calculated .

All Replies
Frequent Contributor
Posts: 108

## Re: Calculating a rolling sum for the last 30 days

Hope this helps :)

data ydelser;
input PERSONID\$ TOTAL;
datalines;
A 10
A 12
A 14
A 17
B 30
B 30
B 40
B 50
;
Data newdata;
set ydelser;
by PERSONID notsorted;
if first.PERSONID then TOTALsum = 0;
TOTALsum+TOTAL;
run;
Proc Print data = newdata;
run;
New Contributor
Posts: 4

## Re: Calculating a rolling sum for the last 30 days

Thanks!

I've forgot to mention that I have hundreds of rows. Do I have so specify all in the datalines?

Frequent Contributor
Posts: 108

## Re: Calculating a rolling sum for the last 30 days

If you need only PERSONID\$ TOTAL and the new variable "TOTALsum" then you don't have to ..
You use Proc Import to import the file ..so you dont have to write anything datelines .

Here is the link ..for Proc Import

https://support.sas.com/documentation/cdl/en/proc/61895/HTML/default/viewer.htm#a000332605.htm
Solution
‎11-26-2015 01:54 AM
Super User
Posts: 9,599

## Re: Calculating a rolling sum for the last 30 days

Here is code which will do a rolling 30 day sum for you.  The datalines or cards is just used to create the test data (I have left in to show you).  I do not understand why the fourth A row in your example out data is 22 though, that doesn't make sense with the logic.

data have;
informat date ddmmyy10.;
format date date9.;
input person_id \$ date total;
datalines;
A 08-01-2014 10
A 13-01-2014 12
A 23-01-2014 11
A 12-02-2014 17
;
run;

proc sql;
create table WANT as
select  PERSON_ID,
DATE,
TOTAL,
sum(TMP) as ROLLING_SUM
from    (
select  A.*,
B.TOTAL as TMP
from    HAVE A
left join HAVE B
on      A.PERSON_ID=B.PERSON_ID
and     A.DATE-30 <= B.DATE <= A.DATE
)
group by PERSON_ID,
DATE,
TOTAL;
quit;

What this does is join all records where date is within 30 days to the original data, this is the subquery.  That data is then grouped by the original variables and a sum calculated .

New Contributor
Posts: 4

## Re: Calculating a rolling sum for the last 30 days

You are right, the fourth row should have had the value 40. Sorry for the mistake.

Frequent Contributor
Posts: 132

## Re: Calculating a rolling sum for the last 30 days

This solution is pretty much the same as @pearsoninst provided; however, it includes the DATE variable and will keep only the last 30 days from the current date.  If you want the last 30 days of a different date than the current, just substitute whatever date you want in place of the today().

data ydelser;
input PERSON_ID\$ DATE ddmmyy10. TOTAL;
datalines;
A 08-01-2014 10
A 13-01-2014 12
A 23-01-2014 14
A 12-02-2014 17
B 07-01-2014 30
B 17-01-2014 30
B 25-02-2014 40
B 03-03-2014 50
;
run;

proc sort data=ydelser;
by PERSON_ID;
run;

data ydel_sum;
set ydelser;
by PERSON_ID;
Where DATE>=intnx("DAY",today(),-30);
If First.PERSON_ID then TOTALsum=0;
TOTALsum+TOTAL;
run;
New Contributor
Posts: 4

## Re: Calculating a rolling sum for the last 30 days

Thank you for all your replies! Both solutions worked, and I have accepted the one that I used as the solution.

🔒 This topic is solved and locked.