topic Re: please help simplify some codes in SAS Programming

please help simplify some codes

fengyuwuzu — Thu, 03 Mar 2016 18:30:22 GMT

my data has ID and stores where they put orders. I want to calculate how many stores one ID has put orders with, and have several steps code, and wonder if someone can help imporve it by making it shorter.

/* first, only select ID and Store columns to make data size smaller */
proc sql;
create table table1 as
select  ID, store
from mydata;
quit;

/* count how many stores one ID has put orders with */
/* this step, in the output data, the same ID listed n times if n stores */
proc sql;
create table table2 as
select ID, store, count(*) as how_many_stores_per_ID
from table1
group by ID;
quit;

/* to remove duplicate rows for the same ID */
proc sql;
create table table3 as
select distinct ID, how_many_stores_per_ID
from table2;
quit;


/* to verify if the sum of all counts equal to the total rows of mydata */
proc means data=table3 n sum ;
var how_many_stores_per_ID;
output  n=count sum=total1;
run;

I tested, they all worked, and the sum(how_many_stores_per_ID) equals total rows of mydata, which means it is right.

Re: please help simplify some codes

fengyuwuzu — Thu, 03 Mar 2016 18:44:41 GMT

I was thinking to do like below, which gave different results. which one is correct, the above moethod or the one below? Many thanks.

proc sql;
create table table4 as
select ID, store, count(*) as how_many_stores_per_ID
from table1
group by ID, store;
quit;

Re: please help simplify some codes

Reeza — Thu, 03 Mar 2016 18:59:14 GMT

Post sample data.

Test it manually with a small subset of data. You know what the answer should be, so you'll know which code is correct.

Re: please help simplify some codes

dcruik — Thu, 03 Mar 2016 19:03:09 GMT

You skipped the step where you removed duplicate rows for ID by grouping by ID and Store. Also, you can do this pulling from your mydata dataset, eliminating the first step. Try the code below:

proc sql;
create table table4 as
	select ID, count(store) as how_many_stores_per_ID
from mydata
group by ID;
quit;

Re: please help simplify some codes

fengyuwuzu — Fri, 04 Mar 2016 15:22:15 GMT

Re: please help simplify some codes

fengyuwuzu — Thu, 03 Mar 2016 19:10:06 GMT

I am confused with: should I use group by ID or
group by ID, store?

Re: please help simplify some codes

dcruik — Thu, 03 Mar 2016 19:12:38 GMT

If you use group by ID, store then you will get duplicate ID's if they are related to more than one store. In your original code, you grouped by ID, store; however, you then proceeded to eliminate duplicate IDs with your select distinct statement. You can group just by ID, and that will give you a count of all the stores for each unique ID.

Re: please help simplify some codes

LinusH — Thu, 03 Mar 2016 19:18:51 GMT

One step should be enough:

select id, count(distinct store)

Re: please help simplify some codes

fengyuwuzu — Thu, 03 Mar 2016 19:24:39 GMT

and use group by id?

Re: please help simplify some codes

LinusH — Thu, 03 Mar 2016 19:26:17 GMT

Yep

Re: please help simplify some codes

LinusH — Fri, 04 Mar 2016 14:32:04 GMT

Did the count(distinct...) work for you?

If yes, mark as solved.

Re: please help simplify some codes

fengyuwuzu — Fri, 04 Mar 2016 15:22:49 GMT

yes, it is the simplest code.
Thanks