Simple idea:  test if the range is 0 or not.  If it is 0, then all non-missing values are the same.  If not, there is spread amongst the values.  Try this: data same different;      set have;      if range(of v1-v5) = 0 then output same;      else output different; run; Note that if there are missing values in the arguments, this WON'T detect its presence.  You'll need additional logic to handle. Message was edited by: Daymond Ling ... View more

1) Since the dataset is sorted, using first.indvid and last.indvid is easier and cleaner than using lag(). 2) the total variable needs to be retained across observations via "retain total;" or using the "a + b;" sum statement. You're probably looking for something like: data claims2 claimsmth1;      set claims1;      by indvid yrmth;      output claims2;                             /* replication of claims1 */      if first.indvid then total = 0;      total = total + paid;     retain total 0;  /*  or you can use "total + paid;" which has an implied retain  */      if total>=100 then output claimsmth1; run; ... View more

*.sas7bitm are SAS item store files used to store various things under SAS's management like registry, ODS graphics, PDF files... etc.  They are not SAS datasets that store data.  If you have SAS datasets, usually one uses SAS to export them to different formats like CSV, Excel. Why do you not want to use SAS to export things?  I'm not aware of general utilities that convert SAS datasets, though there are other statistics software packages that can import SAS datasets into their environment. ... View more

Rapid Predictive Miner, if I understand it correctly, is meant to simplify the model building process by using pre-defined modeling process flow diagrams.  If you have relatively standard problems that RPM process flow can solve well, then it lets you very quickly run some models and see some results. If you need to tweak the process flow or do something fancier, then you need to drop into Enterprise Miner where you drive the entire modeling process, including putting together the process flow the way you want. I think of it as auto-pilot/cruise control versus taking the wheel yourself.  There are times when you'd want one or the other, and the flexibility is there.  RPM isn't necessarily leaving anything out per se. ... View more

proc summary data=sasuser.marchflights missing;     *--- missing option outputs missing class values as well;      class date;                                    *--- you can attach formats to dates to get monthly, qtrly summaries easily;      var x1 x2 x3;      output out=mtsudat.marchflightsummary(drop=_:)     /*  drop _type_ _freq_  */           sum=x1sum x2sum x3sum           n=x1n x2n x3n           std(x1)=x1std      ; run; proc summary by default prints nothing, so maxdec=2 has no effect, nor does the mean option.  They are meaningful for PROC MEANS printed output. ... View more

PC SAS 9.1.3, no SAS/OR (otherwise PROC CLP constraint programming would be fun to play with in addition to PROC OPTMODEL).  Can't vouch for efficiency, but it works. %let r1= 100; %let r2= 316;    data soln(keep=c1-c3 dUsedOnce);     do i = &r1 to &r2;                              *--- generate combinations;         c1 = put( i*i, z5. );         do j = i+1 to &r2;             c2 = put( j*j, z5. );             do k = j+1 to &r2;                      *--- using i+1 and j+1 generates lower triangular only;                 c3 = put( k*k, z5. );                 link Chkit;                         *--- check this combination;             end;         end;     end; return; Chkit:     length c1-c3 $ 5 str15 $ 15;     array digit[10] $ d1-d10 ('1','2','3','4','5','6','7','8','9','0');     array used[10]    u1-u10;     array freq[10]    f1-f10;     str15 = cat( c1, c2, c3 );     do d = 1 to 10;                                 *--- calc freq and used indicator;         freq = countc( str15, digit );         used = ( freq > 0 );         if freq = 1 then dUsedOnce = digit ;     end;     if sum(of used ) ~= 5 then return;            *--- use 5 different digits;     do d = 1 to 10;         if used then do;                         *--- if d is used;             if freq = d then return;             *--- not used its own # of times;             if used[freq ] = 0 then return;       *--- freq is used;             do p = 1 to d-1;                        *--- if p is used, must be different # of times;                 if used > 0 then if freq = freq then return;             end;         end;     end;     output; return; run; proc sql;     select * from soln group by dUsedOnce having count(*)=1; quit; Hats off to PROC FCMP, very nice approach.  Never used it, so I learned something valuable out of this.:smileygrin: Message was edited by: DLing (minor cleanup) ... View more

You're not being dense, you raised a very good point.  You got me on that one, I don't understand why SAS did it that way. As you said, multiple attrib statements, last one wins, no compile time messages generated.  But a redundant declare hash that is syntactically identical does generate an error message. Perhaps ask SAS to "fix" it, or help us understand the situations where multiple declare hash would causes issues. Message was edited by: DLing to fix grammar. ... View more

For Windows PC, WORK usually uses the local hard drive as opposed to LAN or other network storage.  You can point them anywhere you like as long as you have read/write access. In UNIX environments, it is not uncommon to see a central directory and/or filesystem that's built for speed only being used for WORK.  Since these SAS datasets only last for the duration of a SAS job, there's no need for mirroring or stripping. ... View more