Reducing the Conservativeness of Fully Sequential Indifference-Zone Procedures

In this technical note, we study three sources of conservativeness in fully sequential indifference-zone procedures and quantify, by experiments, the impact of each source, in terms of the number of observations, to identify which source is critical. Then we propose new asymptotically valid procedures that lessen conservativeness by mean update with or without variance update. Experimental results show that meaningful improvement on the efficiency is achieved with the new procedures.