topic Re: programming logic in SAS Programming

programming logic

Chandanat17 — Mon, 11 Sep 2023 13:57:07 GMT

There is no Adverse Event reported on Adverse Event Form where
AE.AEDIS = YES

If atleast in any of the logline aedis=yes then it should not fetch in the output. Can anyone suggest a solution?

Re: programming logic

andreas_lds — Mon, 11 Sep 2023 14:06:06 GMT

Please post data in usable form so that we have something to work with. Also show what you expect as output.

Re: programming logic

Chandanat17 — Mon, 11 Sep 2023 14:14:22 GMT

logic: DS.DSDECOD = Adverse Event
AND
There is no Adverse Event reported on Adverse Event Form where
AE.AEDIS = YES

subject aedis dsdecod
1001 Yes Adverse event
1001 blank Adverse event

1001 no Adverse event

Like above example, a subject has 3 loglines and in one of the logline it is present as aedis=yes , so this subject should not come in output.
code for this ??

Re: programming logic

mkeintz — Mon, 11 Sep 2023 14:40:32 GMT

In the absence of sample data in the form of a working data step, the code below is untested.

Assuming your data is in a dataset named HAVE and the data is sorted by subject, then:

data want;
  merge have (where=(dsdecod='Adverse event') in=unwanted)
        have ;
  by subject;
  if unwanted=0;
run;

Re: programming logic

ballardw — Mon, 11 Sep 2023 15:35:53 GMT

Is this one of those questions where ALL of the values for single person has to be considered?

If so you need to state such before getting into any coding. SAS data steps where any such code is likely to be used processes data sets one observation at a time. If something has to consider more than one observation for a 'correct' answer than that criteria has to be explicitly stated and described properly to implement.

Re: programming logic

andreas_lds — Tue, 12 Sep 2023 10:48:54 GMT

When posting data make sure that

you post the data as working data step code
and that all cases that exist in your real data are present at least once.

Idea to solve the problem:

create a dataset containing all obs with "Adverse event"
use that dataset to create another one having only those obs with aedis = "Yes"
finally something like this:

proc sql noprint;
  create table work.want as
    select * from work.adverse_events
     where subject not in (select subject from work.aedis)
  ;
quit;

Code is, of course, untested. Using a single step with two dow-loops seems to be the better approach, as less iteration are required.