Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We show that confidence regions covering the identified set may be preferable to confidence regions covering each of its points in robust control applications.