topic Re: Cumulative sum with time limitation in SAS Programming

Cumulative sum with time limitation

Lipty — Wed, 22 Nov 2017 07:36:53 GMT

Sorry for asking a stupid question.

I have two data sets: A: with ID and date B: with ID newsnumber date. I would like to make data set C which includes the cumulative newsnumber by ID for the past 365 days prior to the date in data A.

Thank you very much for your help.

Re: Cumulative sum with time limitation

PeterClemmensen — Wed, 22 Nov 2017 07:44:05 GMT

Post some sample of your data and what your desired output looks like 🙂

Re: Cumulative sum with time limitation

Kurt_Bremser — Wed, 22 Nov 2017 08:24:42 GMT

Use the macro provided in https://communities.sas.com/t5/SAS-Communities-Library/How-to-create-a-data-step-version-of-your-data-AKA-generate/ta-p/258712 to convert your datasets to data steps, or write your own data steps that create example data.

Post the resulting code according to https://communities.sas.com/t5/help/faqpage/faq-category-id/posting?nobounce (How can I add SAS syntax to a post?)

Re: Cumulative sum with time limitation

Lipty — Wed, 22 Nov 2017 12:45:58 GMT

Data set A:

ID Date

1 1998/07/21

2 2000/08/21

4 2001/04/21

6 1999/08/21

7 2000/06/21

9 2010/08/21

Data set B:

ID Date Newsnumber

....

4 2000/04/22

4 2000/04/23

....

4 2000/06/22

4 2000/06/23

4 2000/06/24

...

4 2001/04/20

4 2001/04/21

...

For Data set C, let's take firm 4 for an example, I would like to get the cumulative sum of Newsnumber for the past 365 days before 2001/04/21.

Thanks~~~~ ^_^

Re: Cumulative sum with time limitation

Reeza — Wed, 22 Nov 2017 16:16:03 GMT

@Lipty Do you have SAS/ETS? If so, look into PROC EXPAND

Re: Cumulative sum with time limitation

Lipty — Wed, 22 Nov 2017 16:23:57 GMT

Yes, I have it and read it last night. But didn't know how to solve for the problem. Thank you

Re: Cumulative sum with time limitation

Reeza — Wed, 22 Nov 2017 16:24:32 GMT

Post what you've tried.

Re: Cumulative sum with time limitation

mkeintz — Tue, 27 Nov 2018 21:55:48 GMT

I know this is a stale thread, but it was on my "recommended by SAS" box, so I took a look. Here's a solution:

data want;
  merge A (rename=(date=adate))
        B ;
  by id ;
  if first.id then cumsum=0;
  if adate-365 <= date <=adate then cumsum+newsnumber;
  if last.id;
run;

Notes:

This program assumes that dataset B is sorted by ID/date, and dataset A is sorted by ID.
It also assumes that A never has more than 1 record per ID.
The program assumes that the user want one record per id, even when cumsum=0.
If you want to account for leap years, then the "if adate-365 <= date ..." statement should be changed to:
if intnx('year',adate,-1,'s') <=date <=adate then cumsum+newsnumber;

The "trick" here is to merge by id, NOT by id date. This results in the dataset A record merged with all the dataset B records for the same id, even though the date in A does not precede all the dates in B. Of course you have to rename date in A, so as to not collide with date values from B. If there is no obs in A for a given id, then adate=.