Finding Earliest begin date and latest end date in a non continuous dataset

[sneha-25]

I have a dataset like this:
ID Start date End date
A 2015-01-01 2015-12-31
A 2016-01-01 2016-12-31
A 2017-01-03. 2017-12-31
B 2020-01-01 2020-12-31
B. 2022-01-01 2020-12-31

I want my final dataset to be
ID Start date End date
A 2015-01-01 2016-12-31
B 2020-01-01 2020-12-31

It should be the earliest begin date and latest end date from the continuous coverage. If there is a break in the next begin date then it should not consider the rows after that for that particular ID.

[mkeintz]

What if you have more than one obs at earliest start or latest end?

And also, your last obs has start of 2022-01-01 and end of 2020-12-31.  Doesn't make sense.  In fact, that end is an actual tie between the last two obs.

Finally, is your data sorted by start?  Your sample is.

[Tom · Accepted solution]

You could just use normal code for collapsing to continuous periods.

So perhaps something like this:

data have;
  input ID $ (start end) (:yymmdd.);
  format start end yymmdd10.;
cards;
A 2015-01-01 2015-12-31
A 2016-01-01 2016-12-31
A 2017-01-03 2017-12-31
B 2020-01-01 2020-12-31
B 2022-01-01 2022-12-31
;

data want;
 set have ;
 by id start end;
 if first.id then do;
    period=0;
    lagend = start-2;
 end;
 retain period lagend min max;
 format lagend min max yymmdd10.;
 if 1<(start-lagend) then do;
   if period then output;
   period+1;
   min=start;
   max=end;
 end;
 max=max(max,end);
 if last.id then output;
 lagend=end;
 drop start end lagend;
run;

Result

OBS    ID    period           min           max

 1     A        1      2015-01-01    2016-12-31
 2     A        2      2017-01-03    2017-12-31
 3     B        1      2020-01-01    2020-12-31
 4     B        2      2022-01-01    2022-12-31

And just modify it to only OUTPUT when PERIOD is 1.

data want;
 set have ;
 by id start end;
 if first.id then do;
    period=0;
    lagend = start-2;
 end;
 retain period lagend min max;
 format lagend min max yymmdd10.;
 if 1<(start-lagend) then do;
   if period=1 then output;
   period+1;
   min=start;
   max=end;
 end;
 max=max(max,end);
 if last.id and period=1 then output;
 lagend=end;
 drop start end lagend;
run;

Result

OBS    ID    period           min           max

 1     A        1      2015-01-01    2016-12-31
 2     B        1      2020-01-01    2020-12-31

[Ksharp]

data have;
input ID $ Startdate :yymmdd12. Enddate :yymmdd12.;
format Startdate  Enddate :yymmdd10.;
cards;
A 2015-01-01 2015-12-31
A 2016-01-01 2016-12-31
A 2017-01-03 2017-12-31
B 2020-01-01 2020-12-31
B 2022-01-01 2020-12-31
;

data temp;
 set have;
 do date=Startdate to  Enddate;
   output;
 end;
format date yymmdd10.;
keep id date;
run;
proc sort data=temp out=temp2 nodupkey;
by id date;
run;
data temp3;
 set temp2;
 by id;
 if first.id then group=0;
 if first.id or dif(date) ne 1 then group+1;

 if group=1;
run;
proc summary data=temp3;
by id;
var date;
output out=want(drop=_:) min=Startdate max=Enddate;
run;