The quantiles in this case (5) are quintiles. Investigate      Proc Ranks groups =  5 data = have  out = want /*descending ? */;           Var income ;           Ranks income_group ;      Run ; [untested] This will number your income values in groups labelled 0 - 4 (4 highest, use descending to reverse the numbering).  Add 1 to income_group to get labels 1 - 5.  Multiply by 20 if you want labels like 20, 40 60, 80. Richard ... View more

@mahler.ji It would have been helpful to have used the term 'quartile' in your original post rather than stating it in terms of ranking.  Proc univariate will provide quartile information, but so will proc summary / proc means, probably easier to use in this context.  However, Proc Univariate data = have ;  /* have is your existing dataset     */      By year ;                         /* you can use Class instead of By if your data is not sorted and not too big */      var     sales ;      Output out = quartiles     q1     =     salesq1                                    median     =     salesq2                                         q3       =     salesq3                                         ; Run ; Now since the quartliles are in the same row the easiest way to proceed is with datasteps, assuming your have data is at least sorted by year.  First adjust the year value in the quartiles to allow matching for the previous year Data quartiles_adj ;      By year ;      Year + 1 ; Run ; Data want ;      merge     have                     quartiles_adj                     ;      By      year ;      Retain firstyear ;      if     _N_     =     1     then firstyear = year ;      Select ;           When (year = firstyear)    Prev_quartile = . ;           When (sales >= salesq1) Prev_quartile = 1 ;           When (sales >= salesq2) Prev_quartile = 2 ;           When (sales >= salesq3) Prev_quartile = 3 ;           Otherwise                       Prev_quartile = 4 ;      End ;      Drop salesq1 - salesq3  firstyear; Run ; [this code is untested] Using SQL would avoid the first datastep so if you are more familiar with SQL it would look like this Proc SQL ;      Create table want as           Select h.*                ,     q.sales:                ,     case                          when sales >= salesq1 then 1                          when sales >= salesq2 then 2                          when sales >= salesq3 then 3                          else     4                     end     as     Prev_quartile           From     have          h                left join                ,       quartiles     q                on     h.year     =     (q.year - 1)           Order     by                         id                   ,     year           ; Quit ; [also untested] Richard ... View more

This is a question about how long data items in a SAS dataset can be displayed in multiple lines.  It is not a question of how the data is stored 'under the covers' because it makes no sense in that context, either for SAS or Excel. I doubt whether you can make the SAS data viewer behave in the way you describe but Proc report (and tabulate) can wrap data within fixed width cells.  An example from the SAS documentation: proc format ;    value questionf    1='This is an example of a description that is too long so it wraps down to the next line'; run;    data a;    input question rater $ avg;    datalines; 1 manager 3 1 staff 2 1 peer 3 ; run;    ods listing close; ods escapechar='^'; options nodate nonumber orientation=portrait; proc report data=a nowd spanrows;    column question rater avg;    define question / order                      format=questionf.                      style(column)={cellwidth=2.2in};    define rater / display;    define avg   / display; run; (this comes from 9.2 documentation that describes an undesired side effect with ODS PDF: 45549 - Wrapped text strings in the first column cause unwanted white space in subsequent rows in PROC REPORT and PROC TABULATE output routed to ODS PDF) Richard ... View more

I think you can meet this somewhat odd requirement by first ranking all ids within each year (proc rank with year as a by variable).  Select the rank option to give the top rank (=1) to the best performer.  Then use SQL to join back the ranked data with the original dataset, where rank.year = (data.year - 1) and select max(rank.rank) where rank.sales >= data.sales (for each id in the chosen year, select the maximum rank [of all ids] in the previous year where the maximum value of last year's sales is greater or equal to the sale for that id in the chosen year).  Someone with access to SAS this weekend should be able to code it. Richard ... View more

The /Q option on the DOS dir command gives the owner of the file.  You could use X or equivalent to issue the DOS command to write the contents to a text file, then import the file into SAS.  Something like dir [drive:][path] /Q /N /S > c:\temp\File_List.txt where /N puts the filename last (easier to read into SAS) and /S allows subdirectories to be listed However, if you use /S you will have to write code to extract the subdirectory information from the header for each directory. Richard ... View more

Treating weight as an ordinal varaible would mean assigning ordered categories of weight that are not (necessarily) linear functions of weight. My suggestion would be to use Proc Rank to split the population into deciles and assign each one a category number.  Other strategies might give somewhat different results.  I doubt there is one "right " way to do this. Richard ... View more