New assumption-lean methods for cluster-randomized trials enable estimation of nonparametric intent-to-treat effects as well as network effects among compilers.

Menu