First, you second record should have  start_time-end_time=180, not 280.   The "trick" here is to read the next start_time in tandom with the current end_time to determine whether the record-in-hand is the last one within 300:   data have; input Ids start_time end_time diff; datalines; 1 14280 16820 . 1 17000 19000 180 /* was 280 */ 1 19100 19200 100 2 18000 18500 . 2 18900 19000 400 3 15000 15500 . 3 15900 16000 400 3 16200 16300 200 3 16400 16500 100 4 18000 19000 . 5 15000 16000 . 5 16300 17000 300 ; data want (drop=_:); do until (last.ids=1 or _next_strt>end_time+300); set have (keep=ids); by ids; merge have have (firstobs=2 keep=start_time rename=(start_time=_next_strt)); if _link_strt=. then _link_strt=start_time; end; start_time=_link_strt; run;   Editted additional note:  Note you don't need to establish a variable to contain start_time-lag(end_time). ... View more

@wentao wrote: That's my code.   But the first row of results still have headers for every variables. And all values are missing quotation marks. I see nothing in your post about not wanting the header row only that you couldn't use the putnames statement.   if _n_ > 1 then put _infile_; will write other than  the header rows.   It appears that you may have destroyed your orginal file by using proc export to write over it. Proc Exported removed any quotes that did not need quotes.. Quotes around a single word or phrase with no embedded commas in CSV are not needed and were likely replaced by Export, not by the PUT _INFILE_ which would keep the quotes. Then rereading the file that had the quotes removed means they are not there to write with a PUT statement.   Also, did you open that CSV with Excel or a text editor such as Wordpad or Notepad? Excel will not show the quotes generally. And may remove those that are there if you save the file.   Your data _null_ program does not need the SET statement. ... View more