For an outcome vector, the Stephenson score of a unit at ascending rank j is
choose(j - 1, r - 1). Ties take average ranks. Larger r concentrates weight
on the top ranks; r=2 reduces to the (rank-1) Wilcoxon score.
Arguments
- y
numeric vector (a single block's outcomes).
- r
the Stephenson tuning value, a whole number;
r=2is the Wilcoxon (rank - 1) score andr=1gives a constant (uninformative) score. Must be a whole number: a fractionalrwould silently round insidechoose().