you don't need to generate 2^n-1 datasets, just let PROC MEANS takes care of it.in the output dataset, the combination of x_i1, xi_2, ..., and x_ik will be identified by _TYPE_=2^{i1-1}+2^{i2-1}+...+2^{x_ik-1} in the output dataset xxx, where 1<=i1<i2<...<ik<=8.

%let libname=sashelp;
%let dsname=class;

data test;
    set sashelp.vcolumn end=last;
        where upcase(libname)=upcase("&libname") and upcase(memname)=upcase("&dsname");
        call symputx(cats("var",_n_),name);
        if last then call symputx("num",_n_);
run;

data test1;
    length code $1000;
    do i=1 to &num;
           code=catx(" ", code, "do var"||strip(i)||"= '           ', "||'"&&var'||strip(i)||'";');
    end;
        code="data test2; "||strip(code)|| "output;"||repeat("end;",&num-1)||"run;";
        call execute(code);
    put code=;
run;

data test3;
    set test2 end=last;;
        keepvar=catx(" ",of var1-var&num);
        length code $100;
        retain n 0;
        if ^missing(keepvar) then do;
            n+1;
           code="&dsname"||strip(n)||"(keep="||strip(keepvar)||")";
        end;
        if _n_=1 then call execute("data ");
        call execute(code);
        if last then call execute("; set &libname..&dsname; run;");
run;

/* try this */
data test;                     /*  note: if you have large number of observations */
input x1-x4 var1;        /*  just keep one observation in this dataset */
datalines;
1 1 1 1 1
2 2 2 2 2
3 3 3 3 3
4 4 4 4 4
;
run;

proc summary data=test;
   class x1-x4;
   var var1;
   output out=test1(where=(_type_>0) drop= _freq_)
                       max(var1) = max_value;
run;

data _null_;
   length outvar $30. outname $8.;
   array x x1-x4;
    set test1;
    outvar=' ';         
    outname='out';
    do i = 1 to dim(x);
      if x(i)^=. then do;  outvar = catx(' ', outvar, vname(x(i)));
                                outname = catx('', outname, i);   
                     end;    outname = compress(outname);
   end;  
   call execute('data '||outname||'; set test (keep='||outvar||'); run;');
run;

Thanks a lot for your help~~~

Thank you very much. I will try it!! Thanks again!