Extend the framework to unavailable unbiased welfare estimates

Extend the minimax-regret sequential experimentation framework to settings in which unbiased estimates of welfare effects are unavailable.

Background

The main results assume that researchers observe Gaussian, unbiased estimates of the welfare effect in each experimental period. The authors explicitly identify extending the analysis to settings without such unbiased welfare estimates as a further unresolved research direction, which would broaden the framework to more realistic settings involving biased, indirect, or otherwise unavailable welfare measurements.