It depends on how you create the SUPP-- dataset.
For example, in our own GUI-based SDTM-ETL mapping software, for non-standard-variables (NSVs) that (unfortunately) need to go in a SUPP-- dataset (we call it “banned” to SUPP–), the variable is just treated as a normal variable (i.e. in the normal domain), and “split-off” at the end of the process when the user selects that option, which is usually the case for FDA submissions. In that case there surely is an algorithm, so that one can set RDOMAIN to “derived” (algorithm = picking up the DOMAIN from the original dataset definition) but then indeed one also needs to supply a corresponding “MethodDef” in the define.xml, describing that algorithm.
In case one however “hard-codes” RDOMAIN on an individual case, e.g. having $RDOMAIN=“LB” in the mapping script, then it surely is “assigned”.
Also remark that the difference between “Derived” and “Assigned” can be pretty small.
What software do you use? In modern, GUI-based software, setting the “Origin” and “Source” can be easily done, usually by GUI “wizards”.
