A claim for sequential designs of experiments∗

We recapitulate several hypotheses testing procedures that leave the total sample size open. This drawback is outweighed by typically reduced sample sizes due to the opportu-nity of early stopping. Whereas sequential methods are standards in quality control and especially clinical trials, experimental economists apparently still favor a fixed sample size design. In this paper we illustrate the advantages of sequential methods (e.g. early stopping, power analysis etc.) and conclude that there is much to learn from clinical trials