OK. Plz post the output you are looking for in order to better understand your question.
Where there was overlapped date , one date or obs correspond to many obs after separated it one by one
,like
cluster n temp
1 1 21oct2020
1 2 21oct2020
here N represent the row number(1 is the first row ,2 is the second row)
" please explain how x is concatenated with n and x in the do loop in your code"
do until(last.temp);
set temp;
by cluster temp;
length x $ 40;
x=cats(x,n);
end;
It process one group (cluster and temp are group variables) data. (e.g. do until(last.temp) )
x concated the row number within the same group.
firstly, x= the first n (1),since x is missing value.
secondly, x= the first and second n ( 12 ),since x is RETAIN in the same data step loop.
.........
until the last.temp ,and output the last group value .if data like above (only two obs), the output is
cluster n temp x
1 2 21oct2020 12
Here x identity which obs has 21oct2020 , (the first obs and second one)
Here x also identity the during of overlap date.
cluster n temp x
1 2 21oct2020 12
1 2 22oct2020 12
1 2 23oct2020 12
The last data step could get this:
cluster n temp x start end
1 2 23oct2020 12 21oct2020 23oct2020
data test;
infile datalines delimiter = ',';
input solddt: date9. rxsup dose med_end_dt: date9. cluster ;
format solddt yymmdd10. med_end_dt yymmdd10.;
datalines;
10sep2009, 16, 200, 26sep2009, 1
02oct2009, 2, 45, 04oct2009, 1
06oct2009, 66, 90.91, 11dec2009, 1
12oct2009, 16, 200, 28oct2009, 1
21oct2009, 12, 62.5, 02nov2009, 1
03nov2009, 33, 181.82, 06dec2009, 1
12nov2009, 2, 45, 14nov2009, 1
12nov2009, 16, 200, 28nov2009, 1
04dec2009, 33, 181.82, 06jan2010, 1
14dec2009, 25, 128, 08jan2010, 1
17dec2009, 2, 45, 09dec2009, 1
15jan2010, 5, 45, 20jan2010, 2
18feb2010, 25, 90, 15mar2010, 3
;
run;
data temp;
set test;
n+1;
do temp=solddt to med_end_dt;
output;
end;
drop solddt med_end_dt rxsup;
format temp date9.;
run;
proc sort data=temp;
by cluster temp;
run;
data temp1;
sum_dose=0;
do until(last.temp);
set temp;
by cluster temp;
length x $ 40;
x=cats(x,n);
sum_dose+dose;
end;
drop dose;
run;
data want;
set temp1;
by cluster x notsorted;
retain start;
if first.x then start=temp;
if last.x then do;end=temp;output;end;
format start end date9.;
drop n x temp;
run;
