topic Re: PROC COMPARE: Processing output dataset in SAS Procedures

PROC COMPARE: Processing output dataset

DavidJ — Wed, 15 Oct 2008 15:24:28 GMT

In the following example:
DATA A;
KEY = 100;
Value1 = 200; Value2 = 400;
run;

Data B;
KEY = 100;
Value1 = 200; Value2 = 410;
run;

proc compare BASE=A COMPARE=B CRITERION=2 NOSUMMARY METHOD=ABSOLUTE NOPRINT OUTNOEQUAL OUT=MYDIFFS;
ID KEY;
run;

The dataset MYDIFFS contains 1 record, as follows (and as expected):
_TYPE_ = DIF
_OBS_ = 1
KEY = 100
Value1 = E
Value2 = 10

However if I try and process the MYDIFFS dataset and attempt to change the E to a 0, SAS will throw an error since Value1/2 are numeric. Is there a simple solution to this?

Re: PROC COMPARE: Processing output dataset

data_null__ — Wed, 15 Oct 2008 16:26:21 GMT

The value is actually special missing value E. To refere to this in code use .E

if value1 eq .E then ...

Re: PROC COMPARE: Processing output dataset

DavidJ — Wed, 15 Oct 2008 17:42:09 GMT

15 years of coding SAS and I had never heard of that one. Thanks.

Re: PROC COMPARE: Processing output dataset

Doc_Duke — Wed, 15 Oct 2008 20:26:41 GMT

"Special missing values" have been around since at least version 5.

They are extensively used in certain pockets of analysis (like surveys where you want to code the reason that a value is missing), but many people go through their entire SAS careers without needing them.

I've rarely needed them, but when I did they were a very valuable tool.

Doc

Re: PROC COMPARE: Processing output dataset

data_null__ — Thu, 16 Oct 2008 00:33:54 GMT

You might like the "generic" missing function. You don't need to know the type.

if missing(value1) then...

Also coalesce if nice for a missing to non-missing value conversion.

[pre]
44 data _null_;
45 do y = 1,.e,3;
46 z = coalesce(y,0);
47 put (_all_)(=);
48 end;
49 run;

y=1 z=1
y=E z=0
y=3 z=3
[pre]