Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We consider randomized experiments with two levels of randomization for measuring social interaction effects. We examine how the randomization should be carried out to estimate the coefficients of interest most precisely.