topic How to average up several rows of different group with condition in SAS Programming

How to average up several rows of different group with condition

LisaYIN9309 — Thu, 19 Oct 2017 02:39:54 GMT

I've come across this problem that I want to average up rows that have exact same OrganizationID, Product and ProductLine, make it one row, with summed numerator as new numerator, summed denominator as new denominator, and summed numerator/summed denominator as new rate(AggregareRate).

The average up will only happen if all three column values are the same(OrganizationID, Product and ProductLine). So in the example below, when either product or productline have different values under same organizationID, the rows stay the way it is - only red rows got averaged up.

Here is the input data:

OrganizationID	Product	ProductLine	numerator	denominator	rate
1	Shoes	Commercial	50	200	0.25
1	Shoes	Commercial	60	120	0.5
1	Shoes	Non-Commercial	30	50	0.6
2	Shoes	Commercial	80	160	0.5
2	Shoes	Non-Commericial	90	180	0.5
2	Cloth	Commercial	70	100	0.7
2	Cloth	Commercial	60	300	0.2

The desired output should look like:

OrganizationID	Product	ProductLine	numerator	denominator	AggregateRate
1	Shoes	Commercial	110	320	0.34375
1	Shoes	Non-Commercial	30	50	0.6
2	Shoes	Commercial	80	160	0.5
2	Shoes	Non-Commericial	90	180	0.5
2	Cloth	Commercial	130	400	0.325

Thank you!

Re: How to average up several rows of different group with condition

PGStats — Thu, 19 Oct 2017 03:42:05 GMT

What have you tried so far?

Re: How to average up several rows of different group with condition

Reeza — Thu, 19 Oct 2017 04:00:23 GMT

PROC MEANS, SQL, SUMMARY, UNIVARIATE, or DATA STEP.

Check the PROC MEANS examples for one related to your question.

https://github.com/statgeek/SAS-Tutorials/blob/master/proc_means_basic

If youre having difficulties with a specific issue from here post your code and log.

Re: How to average up several rows of different group with condition

Shmuel — Thu, 19 Oct 2017 05:06:50 GMT

There are many tools to reach the desired output, as @Reeza mentiond.

You can do it in one step (datastep) or in two steps (proc means or proc summary + datastep).

Attached is the code for one datastep, assuming data is already sorted

BY OrganizationID Product and ProductLine;

data want;
 set have;
   by OrganizationID Product ProductLine;
        retain sum_num sum_denom;
       if first.ProductLine then do; 
            sum_num = nemurator;
            sum_denom = denominator;
       end; else do;
            sum_num =sum(sum_num nemurator);
            sum_denom = sum(sum_denom denominator);
       end;

       if last.ProductLine then do; 
          AggregateRate = sum_num / sum_denom;
          output;
      end;
      format AggregateRate 8.5;  /* or any other preffered format */
run;

Re: How to average up several rows of different group with condition

LisaYIN9309 — Fri, 20 Oct 2017 00:03:54 GMT

This is great, thank you.

Re: How to average up several rows of different group with condition

LisaYIN9309 — Fri, 20 Oct 2017 00:04:23 GMT

Thank you for this!