Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Motivated by problems of model comparison using the Bayes factor, the recent literature proposed the use of Hyvärinen factors. We take up model comparison using Hyvärinen factors in the context of stochastic frontier models in the normal-half-normal, normal-truncated-normal, and normal-exponential cases. We illustrate the new techniques in a data set of large U.S. banks.