topic Re: why does this code not work? in New SAS User

why does this code not work?

Jannie_D — Wed, 12 May 2021 10:14:32 GMT

Hi. I have a big dataset on a secure server.

I am cleaning my data, but now I found out that one code does not work and I can not figure out why, maybe you have an suggestion.

The code is:

data total;

set mergetot;

where not missing(dato_diagnose) and not missing(totsssy) or not missing(totlmdb) or not missing(totdrga) or not missing(totdrgh);

run;

I want to only have the observations where there is a value for dato_diagnose and where at least one of totsssy, totlmdb, totdrga and totdrgh has a value.

somehow I still get observations where there is no value for dato_diagnose.

can you find the mistake?

kind regards

Re: why does this code not work?

PaigeMiller — Wed, 12 May 2021 10:39:22 GMT

Perhaps this, which will test to see if there is at least one non-missing among the rest of the variables

where not missing(dato_diagnose) AND (not missing(totsssy) + not missing(totlmdb) + not missing(totdrga) + not missing(totdrgh))>0;

Re: why does this code not work?

Kurt_Bremser — Wed, 12 May 2021 11:04:47 GMT

In Boolean (yes/no) logic, the AND operator takes precedence over OR (like multiplication takes precedence over addition in calculus). This means that your condition will be evaluated like this:

where
  (not missing(dato_diagnose) and not missing(totsssy))
  or not missing(totlmdb)
  or not missing(totdrga)
  or not missing(totdrgh)
;

To make your condition work as intended, you need to force precedence by using brackets:

where
  not missing(dato_diagnose) and (
    not missing(totsssy)
    or not missing(totlmdb)
    or not missing(totdrga)
    or not missing(totdrgh)
  )
;

Note that it is always a great help to make functional code blocks visible by using indentation.