I'm afraid KNN non parametric method is not my area of expertise but if you want some programming help, I may be able to.
So if you play with lag using
by CID QTR OPENDATE;
with OPENDATE being the child group/inner most sub group, with first and last, you will get the hang of it
data test1; set data; by CID QTR OPENDATE; k=lag(balance); if not first.OPENDATE then diff_pct=round(abs((balance-k)/balance)*100); run;
... View more
