Filling missing dates

[Saba1]

Hi

I have a dataset with some missing monthly dates. Please see below:

data have;
input ID DATE: MMDDYY10. PERMNO SHARES;
format DATE MMDDYY10.;
datalines;

1 01/31/2012 10107 100
1 01/31/2012 10145 250
1 01/31/2012 11340 180
1 03/31/2012 10145 450
1 03/31/2012 25267 450
1 04/30/2012 10145 300
2 12/31/2015 40125 1000
2 12/31/2015 31799 850
2 03/31/2016 40125 720
2 03/31/2016 31799 800
;
run;

I want to complete all missing months in a year with all the observations of the previous month.

data want;
input ID DATE: MMDDYY10. PERMNO SHARES;
format DATE MMDDYY10.;
datalines;

1 01/31/2012 10107 100
1 01/31/2012 10145 250
1 01/31/2012 11340 180
1 02/28/2012 10107 100
1 02/28/2012 10145 250
1 02/28/2012 11340 180
1 03/31/2012 10145 450
1 03/31/2012 25267 450
1 04/30/2012 10145 300
2 12/31/2015 40125 1000
2 12/31/2015 31799 850
2 01/31/2016 40125 1000
2 01/31/2016 31799 850
2 02/28/2016 40125 1000
2 02/28/2016 31799 850
2 03/31/2016 40125 720
2 03/31/2016 31799 800
;
run;

Please guide in this regard. Thanks.

[PeterClemmensen]

data want(keep=id dt permno shares);
    declare hash h(dataset:"have", multidata:"Y");
    h.definekey("id", "date");
    h.definedata("permno", "shares");
    h.definedone();
    declare hash h2(multidata:"Y");
    h2.definekey("id", "dt");
    h2.definedata("permno", "shares");
    h2.definedone();
    declare hiter hi2("h2");

    do until (last.id);
        set have;
        by id;
        if first.id then _min=date;
    end;

    do dt=_min by 0 until (dt gt date);
        if h.check(key:id, key:dt) = 0 then do;
            h.reset_dup();h2.clear();
            do while (h.do_over(key:id, key:dt)=0);
                output;
                h2.add();
            end;
        end;
        else do;
            do rc=hi2.first() by 0 while (rc=0);
                output;
		        rc=hi2.next();
	        end;
        end;
        dt=intnx("month", dt, 1, "e");
    end;

    format dt mmddyy10.;
run;

Result:

id  permno  shares   dt
1   10107   100      01/31/2012
1   10145   250      01/31/2012
1   11340   180      01/31/2012
1   10107   100      02/29/2012
1   10145   250      02/29/2012
1   11340   180      02/29/2012
1   10145   450      03/31/2012
1   25267   450      03/31/2012
1   10145   300      04/30/2012
2   40125   1000     12/31/2015
2   31799   850      12/31/2015
2   40125   1000     01/31/2016
2   31799   850      01/31/2016
2   40125   1000     02/29/2016
2   31799   850      02/29/2016
2   40125   720      03/31/2016
2   31799   800      03/31/2016

[Saba1]

Thanks a lot for your assistance.

[Ksharp · Accepted solution]

That was an challenge . 

data have;
input ID DATE: MMDDYY10. PERMNO SHARES;
format DATE MMDDYY10.;
datalines;

1 01/31/2012 10107 100
1 01/31/2012 10145 250
1 01/31/2012 11340 180
1 03/31/2012 10145 450
1 03/31/2012 25267 450
1 04/30/2012 10145 300
2 12/31/2015 40125 1000
2 12/31/2015 31799 850
2 03/31/2016 40125 720
2 03/31/2016 31799 800
;
run;

data temp;
 merge have have(keep=id date firstobs=2 rename=(id=_id date=_date));
run;

data want;
if _n_=1 then do;
if 0 then set have;
declare hash h(multidata:'y');
h.definekey('id','date');
h.definedata('PERMNO','SHARES');
h.definedone();
end;

  set temp;
  by id date;
  if first.id then h.clear();
  h.add();
  output;
  
  if last.date and id=_id then do;
    key=date;
    do i=1 to intck('month',date,_date)-1;
      date=intnx('month',date,1,'e');
	  rc=h.find(key:id,key:key);
	  do while(rc=0);
        output;
		rc=h.find_next(key:id,key:key);
	  end;
	end;
  end;
drop _:  key rc i;
run;

[Saba1]

@Ksharp: Great. The code works perfectly. Thanks.

[Tom]

Assuming that you want to group by ID and PERMNO it will be easier if you sort it that way.

proc sort;
  by id permno date ;
run;

data want ;
  set have ;
  by id permno ;
  set have(firstobs=2 keep=date rename=(date=next_date)) have(obs=1 drop=_all_);
  if last.permno then next_date=date;
  imputed=0;
  do until(date >= next_date);
    output;
    date=intnx('month',date,1,'end');
    imputed=1;
  end;
  drop next_date;
run;

Obs    ID    PERMNO          DATE    SHARES    imputed

  1     1     10107    2012-01-31      100        0
  2     1     10145    2012-01-31      250        0
  3     1     10145    2012-02-29      250        1
  4     1     10145    2012-03-31      450        0
  5     1     10145    2012-04-30      300        0
  6     1     11340    2012-01-31      180        0
  7     1     25267    2012-03-31      450        0
  8     2     31799    2015-12-31      850        0
  9     2     31799    2016-01-31      850        1
 10     2     31799    2016-02-29      850        1
 11     2     31799    2016-03-31      800        0
 12     2     40125    2015-12-31     1000        0
 13     2     40125    2016-01-31     1000        1
 14     2     40125    2016-02-29     1000        1
 15     2     40125    2016-03-31      720        0

Note:  2012 and 2016 are leap years. Also using either of MDY or DMY order for dates will confuse half of your audience.