Apply format with many values to keep zero counts in proc freq

[SarahDew]

I have a working script, see below, to keep zero counts in a proc freq for one variable. This is based on the solution of a former question on this forum.

data test;
	input person $1. first_digit;
	datalines;
A 1
A 2
A 3
A 1
B 2
B 3
B 4
B 5
C 1
C 6
C 7
;
run;

/* Make sure zero counts are kept*/
proc format;
	value digits
		1='1'
		2='2'
		3='3'
		4='4'
		5='5'
		6='6'
		7='7'
		8='8'
		9='9';
quit;

proc summary data=test nway completetypes;
	class first_digit / preloadfmt order=data missing;
	FORMAT	first_digit	digits.;
	output out=first_digit_counts;
	by person;
run;

PROC	FREQ	DATA=first_digit_counts order=data NOPRINT;
	by person;
	TABLES	first_digit	/	MISSING OUT=distribution 
		(RENAME=(PERCENT=OBSERVED));
	weight _freq_ / zeros;
	FORMAT	first_digit	digits.;
RUN;

Now I need to apply the same script, but the possible values are now 100 to 999 instead of 1 to 9. I want to avoid having to type out all numbers in the proc format step (it can be easily done with the help of excel, but it clutters my code). I already have a table at my disposition that has all possible values for first_3digit (100-999) as a numeric variable. Is there a way I can do the proc format step more efficiently?

Some sample data on which it should work:

data test2;
	input person $1. first_3digit;
	datalines;
A 101
A 253
A 336
A 101
B 245
B 245
B 405
B 504
C 101
C 666
C 789
;
run 

[PeterClemmensen · Accepted solution]

One way is to write your format in a data step and read it into PROC FORMAT like this

data fmt;
    retain fmtname "digits";
    do start=100 to 999;
        label=put(start, 3.);
        output;
    end;
run;

proc format library=work cntlin=fmt;
run;

data test2;
 input person $1. first_3digit;
 datalines;
A 101
A 253
A 336
A 101
B 245
B 245
B 405
B 504
C 101
C 666
C 789
;

proc summary data=test2 nway completetypes;
 class first_3digit / preloadfmt order=data missing;
 FORMAT first_3digit digits.;
 output out=first_digit_counts;
run;

PROC FREQ DATA=first_digit_counts order=data NOPRINT;
 TABLES first_3digit / MISSING OUT=distribution 
  (RENAME=(PERCENT=OBSERVED));
 weight _freq_ / zeros;
 FORMAT first_3digit digits.;
RUN;