There is no difference between the SQL and your desired data step as regards to creation of another table. Because you apparently want to reference rows before your event of interest a standard data step (which operated sequentially) is unsuitable.  You could get it to work possibly using an array of lagged values, or potentially with some hash join; or you could sort the original data in descending order so that before becomes after as far as processing is concerned.  However, sorting will require around three times the disk space of the original table. Try the SQL.  It can handle terebytes of data.  If it runs out of space, the same problem is likely to occur with a datastep.  Richard ... View more

Leave hash techniques aside for the time being until you are more advanced. This question can be readily answered using Proc SQL with a left join and a count function Richard ... View more

But if you are not the owner of the file and the permissions are -rw-------  chmod will fail.  You should talk to you IT guys about changing your user profile to create files with the required permissions. Richard ... View more

You need to issue the unix chmod command to change permissions, using the SAS X command or one of it's cousins. From memory it would be something like X chmod 666 filename ; which might have to be remote submitted Richard ... View more

This is likely to be a setting on your remote server, rather than a SAS option.  You need to talk to your server admin - they may be able to extend the login period for your id. Richard ... View more

860 is not half of 8600. You should be able to work out what to do from my code, mostly by leaving out lines of code. Richard ... View more

Do you need 2 different tables? Data one two ;      set have ;      if mod (_N_, 2) = 1 then           do ;                flag = 'ONE' ;                output one ;           end ;      else           do ;                flag = 'TWO' ;                output two ;           end ; Run ; Richard ... View more

I'll do my best.  I don't have access to a functioning version of SAS so you might need to change a few things. Assuming you have uploaded your tab delimited file to      /sso/biconfig/Lev1/SASApp/model1.txt Filename model1 '/sso/biconfig/Lev1/SASApp/model1.txt' ; /*Split into 2 text files */ Filename values '/sso/biconfig/Lev1/SASApp/model1values.txt' ; Filename labels '/sso/biconfig/Lev1/SASApp/model1labels.txt' ; Data _null_ ;      Infile model1 ;      Input ;      File values ;      If _N_ = 2 then delete ;      Put _infile_ ; Run ; Data _null_ ;      Infile model1 ;      Input ;      File labels ;      If _N_ > 2 then delete ;      Put _infile_ ; Run ; /*Now import model1values.txt and model1labels.txt with proc import using the tab method instead of xls */ /*You'll have to figure the code!*/ Proc transpose data = model1labels                          out = labels                          ;      var _character_ ; Run ; Data labels ;      Set labels (keep = _name_ col1) ;      keep dummy . ;         /* null value */      rename col1 = _label_ ; Run ; Proc transpose data = labels                          out = model1 ; Run ; /* Now just add in the values */ data model1 ;      set model1 (obs = 0)            model1values            ;      drop _name_ ; run ; /* Done! hopefully */ OK if you get error messages with the code let me know and I'll try and help Richard Message was edited by: Richard Carson drop statement added - might need to add _label_ and anything else generated by transpose ... View more

If you decide to use the tab delimited approach (safer than a CSV) let me know and i might be able to be more specific with code. Richard ... View more

Once you sort out the file location issue and you can import the data you will find that SAS assigns each column to a character variable because of the second row.  At that point you can create a process to set up the variable labels in a new table using just the second row, and another table to contain the transformed numerical values (firstobs=2), then append the two. Or you might find it easier to create a tab delimited output file from Excel, upload it to Unix and read it twice, once for the first 2 rows to get variable labels and then a second time to import the values. If all your data is numeric the easiest way to apply the labels to variable names is to create a table from the first 2 rows, then transpose it, keeping Col1 which you rename _LABEL_, adding a dummy numerical value in a data or SQL step, and then transposing back. Richard ... View more