topic total and average values across observations using the data step in SAS Procedures

total and average values across observations using the data step

deleted_user — Tue, 11 Jan 2011 19:54:20 GMT

We have a simple SAS program in which we want to loop through the observations twice.
The first time we will add up a column across all the rows and get the total and average value.
For the second loop we will compare the value from the specific row against the total and/or average mentioned above and make a decision to take certain actions for the specific row.
So the questions may be:
1. How do we loop through more than once (using the SET statement) in a data step? Or
2. Can we pass a value (the total/average) from one data step to another data step? That is, one loop through per data step using the SET statement. Or
3. Is there an easy way to obtain the total/average value across all of the observations?

Thank you so much if someone can shed light on this question.

JC

Re: total and average values across observations using the data step

chang_y_chung_hotmail_com — Tue, 11 Jan 2011 20:13:25 GMT

> 1. How do we loop through more than once (using the SET statement) in a data step?

By reading the input dataset more than once using set statement(s).

> 2. Can we pass a value (the total/average) from one data step to another data step?

Yes, we can.

> 3. Is there an easy way to obtain the total/average value across all of the observations?

Yes, there is.

But this is much easier with proc sql, since unlike other sql implementations, it automatically merges back the summary stats. Below assuming no missing values:

proc sql;

select name, age, mean(age) as meanAge

, ifc(age<(calculated meanAge), "below", "at or above") as label

from sashelp.class

where name like "A%";

quit;

/* on lst

Name Age meanAge label

-----------------------------------------

Alfred 14 13.5 at or above

Alice 13 13.5 below

Re: total and average values across observations using the data step

deleted_user — Wed, 12 Jan 2011 15:15:44 GMT

Chang, thank you for the information. Will update or close this record soon.

Regards,