Name: Liu
Firstname: Qing
Title: Research Fellow
Institution: J&J Pharmaceutical Research and Development
Street: 920 Route 202, P.O.Box 300
City: Raritan
Zip-Code: NJ 08869
Country: USA
Phone: (908) 927-3604
Fax: (908) 526-5196
Email: qliu2@prdus.jnj.com
Authors: Qing Liu
Title: Multiplicity and the Choice of alpha-Spending in Two-Stage Designs
Abstract: Consider hypothesis testing where a larger treatment effect size is anticipated whereas a smaller effect size is still of interest. A conventional approach is to use a group sequential design where the smaller effect size is used to determine the total sample size. An alpha-spending function must then be specified such that the trial will stop early if the anticipated effect size is true. For any given alpha-spending function, optimal two-stage designs can be easily found under the anticipated effect size where one has the option of using either the pre-planned or adaptive sample size for the second stage. These optimal designs are compared with a two-stage design with adaptive sample size where no $\alpha$-spending is used. Issues regarding the definition of overall $p$-values are also discussed.