@ballardw wrote: @Hasnan wrote: Hi Nandeep   I would have used the proc freq method, but just an FYI with either method, it would take quite a while to run!   Regards   Approximately 0.22 sec to create a data set with 1,500,000 records randomly assigning 10 postcodes. 0.14 Seconds for proc freq to summarize.   Of course that is my machine and your mileage may vary. But a proc freq for on the order of 1 or 2 million records for a single variable to a data set does not take very much time unless network or other performance issues arise. I would recommend NOPRINT if you only want a data set with proc freq to avoid potentials display time delays building largish html output tables if you have many codes. Hi,   Yes the NOPRINT certainly seems to have sped up the process, and even on my slow machine the Proc Freq was completed within 45 seconds. ... View more

@RW9 wrote: Something like: data want; set have; if lengthn(pc)=6 then do; ... end; else do; ... end; run; You may not need the do blocks, thats just for illustration.  To note, its sometimes better to code little bit each time, so break off part one and part two then combine in another step - just makes reading the code much simpler e.g.: data want (drop=last first); set have; last=substr(pc,lengthn(pc)-2); first=substr(pc,1,lengthn(pc)-3); want=catx(' ',first,last); run; thank you very much for your help and explanation.   Regards   Nandeep     ... View more