Filling up and down missing values with non-missing values

[expertyejin]

Hello SAS community,

I am desperately trying to figure out how to replace missing values with non-missing values based on a given condition. To be more precise, we fill in a missing value if the time difference between its corresponding date and adjacent date is less than a month; and we leave it as missing if it's one month or over apart. For instance, the following dot should be replaced with 1 since the time difference is less than a month: 

09jan2016 1
15jan2016 .

On the other hand, the following dot should be left as missing as the two events happened over a span of more than one month:

21may2016 .
10jul2016 1

Now, here is a subset of the data set I'm dealing with in my workplace:

data my_data;
input ID Birthday date9. tpin;
label tpin="Temporary Personal Identification Number";
cards;
1 31mar1990 .
1 13apr1990 .
1 20apr1990 0
1 04jun1990 .
1 08jun1990 .
1 19jun1990 1
1 01jul1990 .
1 06aug1990 .
1 13aug1990 2
1 22feb1991 .
2 28feb1991 1
2 15apr1991 .
2 20may1991 0
2 25may1991 .
2 01jun1991 .
2 18jun1991 2
;
run;

If you've understood the rules to apply, the final table should look like this after an up and down algorithm, for each ID:

1 31mar1990 0
1 13apr1990 0
1 20apr1990 0
1 04jun1990 1
1 08jun1990 1
1 19jun1990 1
1 01jul1990 1
1 06aug1990 2
1 13aug1990 2
1 22feb1991 .
2 28feb1991 1
2 15apr1991 .
2 20may1991 0
2 25may1991 0
2 01jun1991 2
2 18jun1991 2

I've tried absolutely everything (merge, update, SQL etc.), unfortunately none of them yields the desired results. I'll be really grateful if anyone could help me out!

Thank you so much!

[ballardw]

You don't describe the rule for getting 31Mar1990 and 13Apr1990 to 0.

Also, are your dates actually SAS date values or character? From what you are displaying I would guess character.

When you have SAS date values there are lots of tools to calculate intervals and such.

[expertyejin]

Dates are definitely in numeric (date9.) format. 

And as far as the rule is concerned, the tpin value for 31Mar1990 is assigned 0 because it is less than a month from 13Apr1990, which itself had been assigned a 0 because it is less than a month from 20Apr1990.

[Ksharp · Accepted solution]

data my_data;
input ID Birthday date9. tpin;
format Birthday date9.;
label tpin="Temporary Personal Identification Number";
cards;
1 31mar1990 .
1 13apr1990 .
1 20apr1990 0
1 04jun1990 .
1 08jun1990 .
1 19jun1990 1
1 01jul1990 .
1 06aug1990 .
1 13aug1990 2
1 22feb1991 .
2 28feb1991 1
2 15apr1991 .
2 20may1991 0
2 25may1991 .
2 01jun1991 .
2 18jun1991 2
;
run;
proc sort data=my_data;by id Birthday;run;
data temp;
 set my_data;
 by id;
 dif_month=Birthday gt intnx('month',lag(Birthday),1,'s') ;
 if first.id then dif_month=.;
run;
data temp1;
 set temp;
 by id;
 if first.id or dif_month=1 then group+1;
 drop dif_month;
run;
data temp1;
 set temp1;
 by id group;
 retain temp;
 if first.group then call missing(temp);
 if not missing(tpin) then temp=tpin;
run;
proc sort data=temp1;by id descending Birthday;run;
data want;
 set temp1;
 by id group notsorted;
 retain want;
 if first.group then call missing(want);
 if not missing(temp) then want=temp;
 drop temp ;
run;
proc sort data=want;by id Birthday;run;

[expertyejin]

How amazing! I can't believe it works! Thank you so much @Ksharp !