Sorry that I didn't see that you revised your original code. Yes, Ian's method will work, although it is slightly more efficient to use the # operator to avoid forming the diagonal matrix: do j=1 to p;    x=data-data[,j];    mat=mat+x*(weights[,j]#t(x)); end; (For an explanation, see Converting between correlation and covariance matrices - The DO Loop ) If you determine that the weights are positive, the you can use do j=1 to p;    x=(data-data[,j])#sqrt(t(weights[,j]));    mat=mat + x*x`; end; ... View more

Each comma separated argument can be an expression or a variable list as specified by the word OF. min(1+2,var1-4,var1-var4,of var1-var4); The third argument is var1 minus var4 The fourth is OF the variable list. You can see how omitting the word OF when a variable list intended could be a problem. You could use  "of array reference" which might help to prevent the var1 minus var4 ommitted OF problem 28         data _null_; 29            array var[4] (1:4); 30            x = min(1+2,var1-4,var1-var4,of var1-var4,of var ); 31            put x=; 32            run; x=-3 ... View more

Bernhard in the options used as your session starts add -initcmd "listing off" that ensures that the old output window is closed (but not the new output viewer) when you have all the windows in the arrangement you prefer, issue the command WSAVE ALL that saves window information in your  SASUSER.PROFILE so ensure you look after that resource with proper opening, saving and closing - perhaps occasional backup too good luck peterC ... View more