You want cumulative frequency, from record 1 through current record, of the current value of status. So let's assume you create dataset HAVE with the series of status values.
If you know in advance all the possible values of status, then this is short and simple. It keeps an array of counts with one element per expected value of status. Arrays with the _TEMPORARY_ description will have their values persist from record to record, but will not be stored in the data set - a perfect place to keep running counts. The FIND function return the posiition in the first argument where the second argument is found.:
data have;
input status :$1. @@;
datalines;
Y Y Y X Y X Y
run;
data want;
array counts {2} _temporary_ (2*0);
set have;
index=find('XY',status);
counts{index}+1;
countif=counts{index};
drop index;
run;
If you don't know in advance all the status values in have, then the thing to do is create a hash object, keyed on STATUS and storing the running countif value for each status, as below.
The least intuitive thing about the hash object in this program is that it is declared only once overall (not once per observations). That's why you see object COUNTS declared and defined in the "if _n_=1 then do" group. "_N_=1" in this program is true only when the first observation is current.
The FIND() method retrieves the current value of COUNTIF based on the status value. If the status value is not already in the object, then FiND() fails, and yields a non-zero return code (RC). Hence set countif to zero.
After incrementing COUNTIF, update the hash object with the new countif value.
data want;
set have;
if _n_=1 then do;
declare hash counts();
counts.definekey('status');
counts.definedata('countif');
counts.definedone();
end;
rc=counts.find();
if rc^=0 then countif=0;
countif=countif+1;
counts.replace();
drop rc;
run;
xxx
... View more