topic How to SUM/aggregate dataset of records in SAS Programming

How to SUM/aggregate dataset of records

hermina — Thu, 19 Oct 2017 18:27:54 GMT

Hi,

I am very new to SAS programming (Enterprise Guide) and have tried combining procedures like Proc SQL, Proc Freq as well as macros, but no code worth posting here because they are full of errors. I would greatly appreciate any help!

Input:

I have been given a dataset consisting of 8000 observations (really there are 80million, but that's not important for the context). Sample below.

ID is a variable but is not needed in the output.

For each observation ONE of the obs_1, obs_2 or obs_3 can have the value of 1, OR they are all 0.

Obs	data_period	ID	EUR	riskgrade	obs_1	obs_2	obs_3
1	200801	a	20	3	0	1	0
2	200801	b	30	3	0	1	0
3	200801	c	50	4	0	0	0
4	200802	a	40	3	1	0	0
5	200803	a	20	4	0	0	0
6	200803	c	10	4	1	0	0
…	…	…	…	…	…	…	…
80000	201512	c	7	4	0	0	1

Desired output:

Sample below (made in Excel).

I need one table for EACH riskgrade (in the real dataset there are 20 different, i.e. I want 20 tables).

Each data_period should be assigned to one row.

OBS with riskgrade 3
data_period	SUM_obs_1	SUM_obs_2	SUM_obs_3	# of observations	SUM of EUR during data_period
200801	0	2	0	2	50
200802	1	0	0	1	40
…	…	…	…	…	…
Total (in sample)	2	2	1	3	90

OBS with riskgrade 4
data_period	SUM_obs_1	SUM_obs_2	SUM_obs_3	# of observations	SUM of EUR during data_period
200801	0	0	0	1	50
200803	1	0	0	2	30
…	…	…	…	…	…
201512	0	0	1	1	7
Total (in sample)	1	0	1	4	87

Re: How to SUM/aggregate dataset of records

PaigeMiller — Thu, 19 Oct 2017 19:45:41 GMT

PROC SUMMARY should do this.

Untested code

proc summary data=whatever;
    class riskgrade data_period;
    var eur obs_1 obs_2 obs_3;
    output out=want sum=sum_eur sum_obs_1 sum_obs_2 sum_obs_3 n(eur)=n;
run;

Re: How to SUM/aggregate dataset of records

hermina — Thu, 19 Oct 2017 19:57:35 GMT

@PaigeMiller Wow, many thanks for your quick response!

I tried out your simple yet effective code on my sample data and it works like a charm.

Will give it a go on my real data tomorrow.

Re: How to SUM/aggregate dataset of records

qxiaowen — Tue, 31 Oct 2017 19:54:21 GMT

Hi there,

I have a similar issue to aggregate my data file. I tried your code, but not apply to my data. I wonder if you can help me out as well.

This is my data:

COHORT_YEAR	FG_Status	Needy_Status	Time2Degree	LocalRace	Gender	count
2008	N	N	Not graduated	American Indian or Alaska Nat	F	1
2009	N	N	Within5Years	American Indian or Alaska Nat	M	1
2008	N	Y	4YorLess	American Indian or Alaska Nat	F	1
2010	N	Y	4YorLess	American Indian or Alaska Nat	M	1
2008	N	Y	4YorLess	American Indian or Alaska Nat	F	1
2012	N	Y	4YorLess	American Indian or Alaska Nat	M	1
2008	N	Y	4YorLess	American Indian or Alaska Nat	F	1
2008	N	Y	4YorLess	American Indian or Alaska Nat	M	1

I have 15000 cases and I need to sum/aggregate the Headcount variable given all the 6 categorical variables, which means the output data keeps the categorical variables, and can be used in Excel pivot tables. As I am learning SAS not for long and tried many ways, but none of them works for me. Hope that you can help me out. Thanks in advance.

Xiaowen Qin