Research
Working Papers
-
Best Arm Identification with Knapsacks: Minimax Policies
A resource-constrained decision maker (DM) designs a continuous-time sequential experiment to determine the best choice from an array of treatments. The DM divides her attention between observing the treatments until one of the resources runs out. Under the minimax regret criterion, we characterize the optimal sampling strategy for two treatments in the following settings: (1) when there is a fixed array of resources and (2) when there is a single resource (money) and the DM can stop adaptively. Our analysis relies on a reformulation of the typical optimal stopping problem in which we model diffusions with respect to the cumulative resource expenditure rather than the elapsed time.
-
Designing Persuasive Experiments (joint with Karun Adusumilli)
Incentives in experimental design are often misaligned: experimenters de- sign and finance experiments to seek regulatory approval, while regulators seek to maxi- mize social-welfare. We propose a framework to resolve this conflict, wherein regulators set a minimum expected welfare threshold, and experimenters optimize designs subject to this constraint. It requires no knowledge of experimenters’ private preferences or costs and mitigates strategic Bayesian persuasion. Under normal priors, sampling according to the Neyman-allocation is always optimal, independent of the specific objectives. Fur- thermore, we characterize the optimal stopping-rule. In a numerical study calibrated to historical clinical-trial data, our framework reduces expected sample-sizes by over 48% relative to classical designs that attain the same social welfare.