Here's a perfect example of how PROC Format doesn't follow these rules. If you run the code below (which includes the dataset in the code), you'll see that the second observation with the missing testscore is always treated as missing data, not "lower than the lowest numeric value." It is never coded as a "Failing grade". It is always distinctly different.
However, the last data step and proc print show how the comparison operators in the DATA step use different logic, and thus code the value as a failing grade.
If I'd had my druthers, it would have kept it as a missing value. This is how SAS treats any other assignment statement equation. If I said "Grade = Testscore * 22", if testscore was missing then grade would have been missing. Seems reasonable to me.
I seem to be in the minority here, but I'll die on this hill. 🙂
data test;
infile datalines missover;
input id $ testscore;
datalines;
1 98
2
3 27
4 52
5 68
6 88
7 92
8 78
;
proc format;
value grades
low-59 = 'Failing grade'
60-high = 'Passing grade';
run;
data applyfmt; set test;
format testscore grades.;
run;
proc print data=applyfmt;
run;
proc freq data=applyfmt;
table testscore;
run;
proc freq data=test;
format testscore grades.;
table testscore;
run;
data applycomp; set test;
if testscore le 59 then Grade = 'Failing Grade';
else if testscore gt 59 then Grade = 'Passing Grade';
run;
proc print data=applycomp;
run;
