topic Re: output does not match code in SAS Programming

output does not match code

Jannie_D — Tue, 02 Mar 2021 08:06:17 GMT

Dear all.

I need help.

I am creating population for a big national study. and while creating the last population I encountered a challenge. As I see it my outcome does not match.

I want to end up with a dataset where all contains the variable Dato_diagnose, all contain one of the following: Labsvar_resultid1="03", Labsvar_resultid2="03", Labsvar_resultid3="03", Labsvar_resultid4="03", Labsvar_resultid5="03" and does not contain any of the following: Koloskopi1_dt, Koloskopi2_dt, Koloskopi3_dt, Koloskopi4_dt, Koloskopi5_dt.

my code is this:

data Pop6;

set All1;

where not missing(Dato_diagnose) and Labsvar_resultid1="03" or Labsvar_resultid2="03" or Labsvar_resultid3="03" or Labsvar_resultid4="03" or Labsvar_resultid5="03" and missing(Koloskopi1_dt) and missing(Koloskopi2_dt) and missing(Koloskopi3_dt) and missing(Koloskopi4_dt) and missing(Koloskopi5_dt);

run;

SAS runs the code, but the output contains missing data on Dato_diagnosis and it contains all the Koloskopi1_dt etc.

what have I done wrong?

kind regards

Re: output does not match code

Kurt_Bremser — Tue, 02 Mar 2021 10:34:42 GMT

This is your code, with the logic made visible:

data Pop6;
set All1;
where
  (
    not missing(Dato_diagnose)
    and
    Labsvar_resultid1="03"
  )
  or
  Labsvar_resultid2="03"
  or
  Labsvar_resultid3="03"
  or
  Labsvar_resultid4="03"
  or
  (
    Labsvar_resultid5="03"
    and
    missing(Koloskopi1_dt)
    and
    missing(Koloskopi2_dt)
    and
    missing(Koloskopi3_dt)
    and
    missing(Koloskopi4_dt)
    and
    missing(Koloskopi5_dt)
  )
;
run;

I added parentheses (which do not change the logic!) to illustrate the precedence of AND before OR.

First, the ANDs are evaluated, and the results combined with the ORs

Re: output does not match code

Jannie_D — Tue, 02 Mar 2021 09:06:46 GMT

Thank you so much. Now it works. You are a real hero!

Re: output does not match code

ballardw — Tue, 02 Mar 2021 09:07:26 GMT

Code without data does not show anything about "output does not match code". You would need to provide 1) some starting values for the variables in form of a data step and 2) the expected values that are not matching expectation, again in the form of a data step.

Boolean combinations of AND and OR really require knowing how the order of operations apply results.

You can test whether any of a list of variables has a specific value using the WHICHC, for character values, or WHICHN, for numeric values. The function will return the first position in the list where the value is found or 0 if not found. So "Labsvar_resultid1="03" or Labsvar_resultid2="03" or Labsvar_resultid3="03" or Labsvar_resultid4="03" or Labsvar_resultid5="03" " could be replaced with:

Whichc('03',Labsvar_resultid1,Labsvar_resultid2,Labsvar_resultid3,Labsvar_resultid4,Labsvar_resultid5) > 0

You do not mention if Koloskopi1_dt and related are numeric or character values. If they are numeric you could replace:

missing(Koloskopi1_dt) and missing(Koloskopi2_dt) and missing(Koloskopi3_dt) and missing(Koloskopi4_dt) and missing(Koloskopi5_dt)

with

Nmiss(Koloskopi1_dt, Koloskopi2_dt,Koloskopi3_dt,Koloskopi4_dt,Koloskopi5_dt) = 5

Which would reduce the whole thing to two ANDs:

Where not missing(Dato_diagnose)

and (Whichc('03',Labsvar_resultid1,Labsvar_resultid2,Labsvar_resultid3,Labsvar_resultid4,Labsvar_resultid5) > 0)

and (Nmiss(Koloskopi1_dt, Koloskopi2_dt,Koloskopi3_dt,Koloskopi4_dt,Koloskopi5_dt) = 5)

;

Re: output does not match code

Kurt_Bremser — Tue, 02 Mar 2021 10:36:46 GMT

Note that the code I posted shows the importance of proper code formatting. For this, also study Maxim 12.