This part is just to unitialize the array to zero for each level of the BY variables (ID).  You could leave it out and get missing instead. If works by POKEing a string of repeating ZEROS formated RB8. into the array WCAT.  ADDR is the memory location of element 1 of WCAT and LENGTH is the total memory length of WCAT (8*number of CATs * number of analysis variables). It is effectively do i = 1 to dim(wcat);    wcat=0;    end; It should be faster. Like I said you can leave it out and get the same result with missing for any CAT that does not exist for an ID. Post some sample data and I make the specific code you need. ... View more