> From: Shannon [mailto:shannon_ritter@ALUMNI.ALBANY.EDU] > Suppose you want to perform a similar operation on a group of > variables in the same data set. I usually do this with a > macro do loop. In the current situation I run a proc contents > and then subset based on a substring in the output NAME > variable. This substring then becomes necessary in and after > the macro do loop. The result of the do loop is a data set > that is 1-1 merged with the contents > output(subsetted) data set. The question is should I create > the substring variable before the do loop (which would > involve an extra data step) or after the do loop (in a data > step I need anyway)? In the latter case I actually end up > creating the substring 3 times, once for the first subsetting > (with a where option on the out= data set), once in the macro > do loop (using %substr) and the third time for keeps after > the do loop (by assignment). In the former case I need an > extra data step.

this sounds similar to an algorithm I use.

I recommend creating your necessary variable -- which name is the prefix of the group -- as early as possible which I think, according to your description, 'involves an extra data step'.

In this preprocessing data step you can completely declare your data structure which is always a Good Thing.

Ron Fehd the macro maven CDC Atlanta GA USA RJF2@cdc.gov

Your task is simple: remove the difference between how things should be and how they really are. -- Ashleigh Brilliant pot-shot #4247

... as early as possible.