How to create analysis variable at diff level than input file level?

[Maisha_Huq]

Hi everyone,

I'm wondering if anyone has suggestions for how I could create some meal-level and school-level variables from a food-level file. More specifically,

- Input file is a food item-level file of about 200,000 food items, where each record has variables to indicate the school ID, weekday (num, 1-5), the meal (lunch/breakfast), the food category (five categories), and multiple variables to indicate nutrient contents for that food item. Since these are food-level files, any school-level and menu-level variables repeat for each food within a school or day.  

How would I tell SAS to give me a variable of the minimum of a certain nutrient (ie lowest fat) at a specific school, on a day, during lunch meal, and within one food category.  What kind of functions would I use to get to this school-level variable?

Due to the size and structure of the input file, I'm opting not to do a basic proc transpose at the beginning of this work to convert the food level file to a school - level one. I'm wondering if there s other data/proc? step processing I can do to get to where I need to.

Would you let me know if I should clarify?

Thank you!

[RW9 · Accepted solution]

Its hard to say, post a datastep with some test data,and what you want out.  There is min(), arrays, retain etc. any of which could work.

[Kurt_Bremser]

proc sort
  data=have (
    where=(
      school="school" and
      weekday=X and
      category="category" and
      meal='Lunch'
    )
  )
  out=want
;
by nutrient_content;
run;

The first observation would then be the one with the lowest content for your selection.

[Reeza]

You're looking for summary functions. 

SAS SQL and PROC MEANS are two standard way to obtain descriptive statistics and aggregate data.