I am not sure, but the problem may be that you are using a Unix path convention under Windows. Normally, Windows paths are written with a backslash, and when you refer to a network path in Windows, you will normally use a double backslash first, e.g.: libname mib '\\server\data\user\mib';   ... View more

You can just do it with a single function call, as %sysfunc accepts a format (in this case year4.): %let yr=%sysfunc(date(),year4.); ... View more

Not sure if this is what you want, but it sounds like you want to use SQL instead of merge in order to get a cartesian product, so here is a suggestion: proc sql; create table want as select coalesce(vorige.khiref,huidige.khiref) as khiref, case when vorige.khiref is null or huidige.khiref is null then '6' when knkatd ne knkatd_v then '6' else huidige.knbkty /* or is it vorige.knbkty? */ end as knkbty, /* rest of columns */ from Work.varian1 as vorige full join in.grptabel as huidige on vorige.khiref=huidige.khiref ; ... View more

As @Shmuel suggested, you may be better off with more standard variable names. But if that is possible, rename e.g. '71190 food' to something more similar, such as V791190_food - then you can sum those variable using adm_cost=sum(of V71190:,of V71230:) That approach will not work with variable names that start with numbers.   If you cannot (or will not) rename your variables, here's a possible solution: data want; set have; array nums _numeric_; adm_cost=0; physio_cost=0; do _N_=1 to dim(nums); prefix=substr(vname(nums(_N_)),1,5); select(prefix); when('71190','71230') adm_cost+nums(_N_); when('71340','71350') physio_cost+nums(_N_); otherwise; end; end; run;   ... View more

If you get this data by reading from a file, and you know the file is sorted, you can do it like this: data lb; input usubjid lbdtc $ 16. lbtptnum rbc wbc visit; if _infile_=lag(_infile_) then do; error 'Duplicate record'; delete; end; cards; 101 01-10-2018t10:15 1 1 1 1 86 101 01-10-2018t10:20 2 2 2 1.01 101 01-10-2018t10:20 2 2 2 1.01 ;run; I used ERROR rather than writing a warning, because the ERROR statement will dump all variables in the log. ... View more

Here's how I would do it: %let first_year=2012; %let last_year=2014; proc transpose data=have out=long; var col1-col4; by year pqr notsorted; run; proc sort data=long; by _name_; run; data want; do until(last._name_); set long; by _name_; total+pqr*col1; array years(&first_year:&last_year) year&first_year-year&last_year; years(year)+pqr*col1; end; output; call missing(total,of years(*)); keep _name_ total year&first_year-year&last_year; run; ... View more

Ah, so you want a by variable. I would suggest something like this: %macro Pairwise_Chisq(data=, group=, outcome=, count=,by=); %local byclause bySQL prxid i w; %if %length(&by) %then %do; %let by=%sysfunc(compbl(&by)); %let bySQL=%sysfunc(tranwrd(&by,%str( ),%str(,a.))),; %let prxid=%sysfunc(prxparse(s/(\S+)/ and a.$1=b.$1/)); %let ByClause=%sysfunc(prxchange(&prxid,-1,&by)); %syscall Prxfree(prxid); %put _local_; %end; proc sql; create table groups as select distinct a.&bySQL a.&group as group1,b.&group as group2 from &data a, &data b where a.&group<b.&group order by a.&bySQL a.&group,b.&group; ; quit; data ChiTest; %if %length(&by) %then set &data(keep=&by); ; length Pair_compared $100 Chi_stat DF P_value 8; stop; run; filename tempsas temp; data _null_; set groups; file tempsas; put "proc freq data=&data noprint;" / " output out=CTout PCHI;" / "tables &group*&outcome /chisq cellchi2 nopercent;" ; length wherecls $1000 byvalues $200; if vtype(group1)='N' then wherecls=cats("&group in(",group1,',',group2,')'); else wherecls=cats("&group in ('",group1,"','",group2,"')"); %do i=1 %to %sysfunc(countw(%str( )&by)); %let w=%scan(&by,&i); if vtype(&w)='N' then do; call catt(wherecls," and &w=", &w); call cats(byvalues,&w,','); end; else do; call catt(wherecls," and &w='", &w,"'"); call cats(byvalues,"'",&w,"',"); end; %end; put 'where ' WhereCls ';'; if "&count" ne " " then put " weight &count;"; put 'run;' / 'proc sql;' / " insert into ChiTest(&bySQL Chi_stat,df,P_value,Pair_compared)" / ' select ' byvalues '_PCHI_,DF_PCHI,P_PCHI,"' group1 'vs ' group2 '"' / ' from CTout a' / ' ;' / 'quit;' / ; run; %include tempsas; %mend Pairwise_Chisq; It is not really by variable processing, as it is all done in WHERE clauses, so your data does not need to be sorted. ... View more

Obviously an old macro, and it has an error: What is the TYPE variable doing in the CALL SYMPUT statement (in the Group_pair data step)? Should it not be &Group?   When I ran your code on a dataset, I got no duplicates, the output was nicely ordered by PAIR_COMPARED.   You will have to be more specific about what code you submitted, and what results you got if you need more help.   I did take a closer look at the macro, though. Here's how I would write such a thing: %macro Pairwise_Chisq(data=, group=, outcome=, count=); proc freq data=&data noprint; table &group/out=out_group; run; proc sql; create table groups as select a.&group as group1,b.&group as group2 from out_group a, out_group b where a.&group<b.&group order by 1,2; ; quit; data ChiTest; /* You may want to change the length of the Pair_compared variable */ length Pair_compared $100 Chi_stat DF P_value 8; retain _character_ ' ' _numeric_ .; stop; run; filename tempsas temp; data _null_; set groups; file tempsas; put "proc freq data=&data noprint;" / " output out=CTout PCHI;" / "tables &group*&outcome /chisq cellchi2 nopercent;" ; if vtype(group1)='N' then incls=cats(group1,',',group2); else incls=cats("'",group1,"','",group2,"'"); put " where &group in(" incls ');'; if "&count" ne " " then put " weight &count;"; put 'run;' / 'proc sql;' / ' insert into ChiTest(Chi_stat,df,P_value,Pair_compared)' / ' select _PCHI_,DF_PCHI,P_PCHI,"' group1 'vs ' group2 '"' / ' from CTout' / ' ;' / 'quit;' / ; run; %include tempsas; %mend Pairwise_Chisq; There are several advantages to this version: The TYPE variable error has been fixed (I think). It works with both numeric and character group variables. The COUNT parameter is now optional, it is not clear that you always have a weight variable. It is much easier to look at the code used to create the output, just take a look at the TEMPSAS file after submitting. ... View more