You do it in a data step. data lookup; do year = 2000 to 2013; output; end; run; data nr_of_ind2; merge   nr_of_ind (in=in_noi)   lookup ; by year; if not in_noi then noi = 0; run; ... View more

But that would also mark 2 successive empty flag1 with flag2 = 'Y'. ... View more

I came to the following solution: data comp (keep=id h1 flag2 rename=(h1=time)); set have; by id; retain flag2 'Y'; flag1_l1 = lag(flag1); flag1_l2 = lag2(flag1); time_l1 = lag(time); id_l1 = lag(id); id_l2 = lag2(id); if not first.id then do;   if flag1 = 'Y' and flag1_l1 = 'Y' then do;     if flag1_l2 ne 'Y' and (id_l2 = id or id_l2 = .) then do;       h1 = time_l1;       output;     end;     h1 = time;     output;   end; end; run; data want; merge have comp; by id time; run; ... View more

In the past, I have tried to achieve a consistent data rate of 100 MB/s writing on the SAN, but our workspaces are located on internal disks. Given that your individual files are much larger than ours, and your SAN may also be required to handle the WORKs, I'd go for 500 MB/s. If somebody asks, just tell them that data warehousing IS an I/O intensive task; where DB systems need to be quick in handling small data chunks randomly, DWH's usually go through _all_ the data in a more or less sequential way. ... View more

proc sort data=my_table; by year; run; data nr_of_ind (keep=year noi); set my_table; by year; retain noi; if first.year then noi = 0; if income > 100000 then noi + 1; if last.year then output; run; This would only fail to give you an output if you have no individual entry for a given year at all. Another option would be to create a table with only "year" from 2000 to 2013, and merge/join that back with your initial result and set noi to zero if no record in my_table is present. ... View more

Create an additional variable that contains the year and month (newvar = put(date,yymmn6.). You can then do a proc summary with class userid newvar; ... View more

Then it is a question of the tool that you use to access the remote desktop, maybe their documentation is of help. I'm a UNIX guy and not much help when it comes to Windows servers. ... View more

Once your conditions get complicated, I recommend doing them in a data _null_ step and put the result into a macro variable with call symput('cond_var','Y') or call symput('cond_var','N'), respectively. That way you can easily check if it works by simply doing a %put &cond_var after the data _null_ step. ... View more

If the remote server runs on UNIX, you probably run it through some kind of X windowing environment. X Window servers on PCs usually paste the content of the clipboard when you click the middle mouse button. ... View more

I know from experience with a different setup (SAS on AIX, DB/2 on Linux) that SAS makes use of a local DB/2 client installation (SAS can't simply connect to a DB/2 server all by itself). You need to make sure first that the client instance on the SAS machine can connect to the DB in question, then you can set up the conncetion in SAS. You positively need SAS tech support and your DB administrator to help. It is not trivial. ... View more

I once had the pleasure of attending a presentation at SEUGI in Florence where the ideal setup of disk storage for SAS was the main theme, and I took the following with me: If you have a SAN, it is the responsibility of the SAN manager to provide you with enough throughput; can't do much yourself there. If you work with local disk storage, consider the following: Is an outage in the DWH service caused by a failed disk tolerable? If not, always use some kind of mirroring, RAID 1 preferred (the fastest method for writing), starting with 2 disk adapters across which you place the mirrors. Try to saturate the adapter's maximum throughput with disks. Out of the old SCSI days: adapter supports 320M/sec, disks deliver 40m/sec -> 8 disks per adapter; set up striping so that all disks work in parallel and every access causes all disks (in a logical group) to work. Set up several physically independent volumes and assign those as "special work" libraries. This allows you to do MERGE's where SAS reads infile 1 from disk (group) A, infile 2 from disk (group) B and writes outfile to disk (group) C. You'll be suprised what you can gain with that. Set the UTILLOC so that it points to a different disk (group) than WORK. This prevents disk contention when doing a sort. Most of this applies to magnetic storage, SSD's avoid many of the problems caused by latency and seek times, alleviating the effects of disk contention. ... View more

I strongly recommmend to get in contact with SAS tech support. You situation looks very exotic. ... View more