Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We show how to minimize the probability of misclassifying individuals as being poor or not poor when data on some of their relevant attributes are missing, but an estimate of the population distribution of attributes is available.